JP2005267578A - Dictionary enhancement support system, dictionary enhancement support method, and dictionary enhancement support program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To support a dictionary writer to appropriately and easily judge whether revision of dictionary should be registered or not by reducing a burden in work of the dictionary writer when analyzing influence of revision of dictionary on the basis of differences of results of a lot of analysis. <P>SOLUTION: When this dictionary enhancement support system receives dictionary revision information from an input device, this system generates a temporary registered dictionary by revising dictionary on the basis of the dictionary revision information in relation to the master dictionary. When this system receives input sentences from the input device, the system selects a sentence containing a surface of a subject word of revision of dictionary among the received input sentences, and carries out language processing of the subject word of processing on the basis of the master dictionary and the temporary registered dictionary. In order to analyze influence of the revision of the dictionary, a result of the language processing based on the master dictionary and a result of language processing based on the temporary registered dictionary are compared with each other from various points of view by using an influence analysis table. Furthermore, a result of the analysis and a result of language processing are output, and provided to the dictionary writer. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a dictionary strengthening support system, a dictionary strengthening support method, and a dictionary strengthening support program for supporting a correction work for strengthening a word dictionary used for natural language processing, and in particular, dictionary correction such as word registration by a dictionary operator. The present invention relates to a dictionary strengthening support system, a dictionary strengthening support method, and a dictionary strengthening support program that support registration judgment when performing a search.

  In order to obtain an appropriate analysis process result in a system for analyzing natural language sentences, all the words included in the sentence to be processed and attribute information indicating the grammatical attributes of the words are registered in advance in the word dictionary. Need to be.

  For this reason, the dictionary operator uses a system for correcting the dictionary so that an appropriate analysis result based on the natural language sentence to be processed can be obtained. Dictionary correction such as deletion of attribute and correction of attribute information was performed.

  The dictionary correction operation is performed while the dictionary operator confirms whether or not the dictionary correction adversely affects the context other than the context of interest.

  In conventional dictionary development, in general, in order to examine the influence of dictionary correction, a predetermined amount of example sentences for evaluation are prepared in advance, and the analysis results for the example sentences are compared and evaluated before and after dictionary correction. Then, based on the difference between the comparatively evaluated analysis results, the dictionary operator confirms whether the intended analysis result is obtained or whether side effects are not caused in an unexpected context (for example, non-patent literature) 1).

  In the above confirmation method, if the number of example sentences for evaluation is increased to cover many contexts, the amount of analysis results to be confirmed increases.

  If you are an experienced dictionary worker who has the skill to quickly identify and judge the impact of dictionary correction in many contexts from a large amount of analysis results, the impact of dictionary correction is as expected and the side effects It is possible to determine whether there is any problem relatively quickly, but it is a heavy burden for dictionary workers who are not familiar with the analysis processing system to determine the effects of dictionary correction by checking a large amount of differences. It becomes.

  Conventionally, as a technique for reducing the burden of checking the difference in analysis results of a dictionary worker, a dictionary strengthening support method by presenting analysis result differences of example sentences has been proposed (see, for example, Patent Document 1).

  Patent Document 1 describes that a large amount of analysis result differences are aggregated for each identical morpheme string, and different differences are presented together with the number of appearances.

JP-A-8-287057 Takemoto, Fukushima, “Implementation and Evaluation of Morphological Analysis of Japanese Sentences Containing Spoken Expressions, Proc. Of the 48th Annual Conference of Information Processing Society of Japan”, March 23, 1994, 3rd volume, p. 37-38

  However, in the technique described in Patent Document 1, although the number of times the same difference is viewed by aggregation is reduced to one, a remarkable one is identified from many differences in analysis results in various contexts, and the effect of dictionary correction can be quickly determined. Since it is not easy to judge, there is a problem that the burden of the work of judging the influence of dictionary correction is still large when there are many example sentences for evaluation. This problem is particularly significant for dictionary workers who are not familiar with the analysis processing system and are not skilled in dictionary work.

  As more example sentences are prepared to confirm the effect of dictionary correction in various contexts, the analysis results to be confirmed by the dictionary operator increase. Therefore, the example sentences for evaluation are used to cover many contexts. As the number increases, the burden of determining the influence of dictionary correction becomes even more excessive.

  The present invention solves the above-described problems, reduces the workload of dictionary workers when analyzing the effects of dictionary correction based on many differences in analysis results, and is a dictionary worker who is not skilled in dictionary work. However, it is an object of the present invention to support a dictionary operator so that it can be determined appropriately and easily whether or not dictionary correction should be registered.

  The dictionary strengthening support system according to the present invention is a dictionary strengthening support system (for example, dictionary strengthening support system 100, 110, 120, 130, 140, 150) that supports correction work for strengthening a word dictionary used for natural language processing. Based on master dictionary storage means (for example, master dictionary storage unit 31) for storing a word dictionary in which word information including words is registered as a master dictionary, and dictionary editing information indicating correction instruction contents of the word dictionary, the word A dictionary editing means (for example, dictionary editing means 22) for creating a temporary registration dictionary obtained by editing a dictionary; a temporary registration dictionary storage means (for example, temporary registration dictionary storage unit 32) for storing the temporary registration dictionary created by the dictionary editing means; , Including words that are candidates for correction in the dictionary editing information from candidate processing sentences that are candidates for natural language processing A processing target sentence selection means (for example, processing target sentence selection means 24) for selecting a sentence as a processing target sentence, and a processing target sentence selected by the processing target sentence selection means based on the master dictionary and the temporary registration dictionary. Language processing means for executing language processing (for example, language processing means 23), processing result of natural language processing for the processing target sentence based on the master dictionary by the language processing means, and the processing based on the temporary registration dictionary by the language processing means The result of natural language processing on the target sentence is compared, and the influence of dictionary editing according to the correction instruction content indicated by the dictionary editing information is determined in advance (for example, the “phenomenon seen in the difference” shown in FIG. 21). Dictionary correction influence analysis means (for example, dictionary correction influence analysis means 25) for analyzing each phenomenon) and dictionary correction shadow Analysis presenting means for presenting the analysis result by the analysis means together with the processing result of the natural language processing (e.g., a dictionary modification management unit 21, an output device 4), characterized in that a.

  With the above-described configuration, it is possible to reduce the workload of the dictionary worker when analyzing the effects of dictionary correction based on a large number of analysis result differences, and the dictionary worker is not skilled in dictionary work. However, it is possible to assist the dictionary operator so that it can be determined appropriately and easily whether or not dictionary correction should be registered.

  From the viewpoint of comparative analysis between the processing result of the natural language processing based on the master dictionary and the processing result of the natural language processing based on the temporary registration dictionary (for example, the item “Phenomenon seen in the difference” shown in FIG. 21 is set. An impact analysis table in which each phenomenon) and a numerical value indicating the influence of the estimated dictionary editing (for example, each numerical value set in the item of “dictionary correction influence score” shown in FIG. 21) is associated with each other. An impact analysis table storage unit (for example, an impact analysis table storage unit 33) for storing may be provided, and the dictionary correction impact analysis unit may be configured to perform a comparative analysis according to the impact analysis table (for example, step S106).

  The viewpoint of the comparative analysis is, for example, any one of the phenomena seen in the difference between the processing result of the natural language processing based on the master dictionary and the processing result of the natural language processing based on the temporary registration dictionary. It may be configured to include the presence or absence of a generated word and the presence or absence of a decrease in the number of morphemes in the processing target sentence (for example, see FIG. 21).

  An example sentence storage unit (for example, example sentence storage unit 34) that stores an example sentence prepared in advance as a processing target candidate sentence is provided, and the processing target sentence selection unit performs processing from the processing target candidate sentence including the example sentence stored in the example sentence storage unit. The target sentence may be selected (for example, step S201).

  Example sentence registration means for registering example sentences in the example sentence storage means (for example, example sentence registration means 26), and input sentence reception means for receiving an input sentence input by the correction operator (for example, a part for executing step S103 in the dictionary correction management means 21) The processing target sentence selecting unit selects a processing target sentence from the processing target candidate sentences including the input sentence received by the input sentence receiving unit (for example, step S104), and the example sentence registration unit is selected by the input sentence receiving unit. The accepted input sentence may be registered in the example sentence storage means as an example sentence (for example, step S301).

  Input request presenting means (for example, dictionary correction management means 21, output device 4) that presents an input request for an input sentence as a processing target candidate sentence to a correction worker, and the processing target sentence selecting means includes dictionary editing information. If there is no sentence including the word that is the correction target in the processing target candidate sentence (for example, N in step S401), the non-selection information indicating that there is no processing target candidate sentence that can be selected as the processing target sentence is transmitted. The input request presenting means presents an input request for the input sentence based on the reception of the unselectable information (for example, notifies the dictionary correction management means 21 that the processing target sentence cannot be selected) (for example, step S402). ) May be configured.

  If the input sentence received by the input sentence receiving means is different from the example sentences registered in the example sentence storage means, the example sentence registration means registers the input sentence as an example sentence in the example sentence storage means (for example, in FIG. 8). Step S301) shown in FIG.

  The processing target sentence selecting unit selects a processing target sentence from the processing target candidate sentences including the example sentence stored in the example sentence storage unit and the input sentence received by the input sentence receiving unit (for example, step S104 shown in FIG. 9). , S201), in the example sentence registration means, the analysis result of the dictionary correction influence analysis means for the input sentence selected as the processing target sentence is different from any of the analysis results of the dictionary correction influence analysis means for the example sentence selected as the processing target sentence. In this case, the input sentence may be registered in the example sentence storage means as an example sentence (for example, step S301 shown in FIG. 9).

  Input sentence receiving means (for example, dictionary correction management means, input device 1) for receiving an input sentence input by a correction operator is provided, and the input sentence receiving means accepts input of the natural language processing result of the input sentence together with the input sentence. (For example, step S701), the processing target sentence selection unit selects an input sentence including a word that is a correction target in the dictionary editing information as a processing target sentence, and corresponds to the input sentence selected as the processing target sentence. The natural language processing result is selected as the reference language processing result (for example, step S104 shown in FIG. 10), and the dictionary correction influence analysis unit includes the processing result of the natural language processing on the processing target sentence based on the master dictionary by the language processing unit, The processing result of the natural language processing for the processing target sentence based on the temporary registration dictionary by the language processing means and the processing target sentence selection means 10 is compared with the reference language processing result of the selected processing target sentence, and the influence of dictionary editing according to the correction instruction content indicated by the dictionary editing information is analyzed for a predetermined viewpoint (for example, the step shown in FIG. 10). (S106) The analysis result presenting means may be configured to present the analysis result by the dictionary correction influence analyzing means together with the processing result of the natural language processing and the reference language processing result (for example, step S107).

  An example sentence storage unit (eg, an example sentence storage unit 34) that stores an example sentence prepared in advance as a processing target candidate sentence, the example sentence storage unit stores the example sentence together with the reference language processing result of the example sentence, and the processing target sentence selection unit includes In addition, an example sentence including a word to be corrected in the dictionary editing information may be selected as a processing target sentence, and a reference language processing result corresponding to the example sentence selected as the processing target sentence may be selected. Good.

  Example sentence registration means (for example, example sentence registration means 26) for registering example sentences in the example sentence storage means, reference language processing result registration means (for example, reference language processing result registration means 27) for registering reference language processing results of example sentences in the example sentence storage means, and The example sentence registration unit registers the input sentence received by the input sentence receiving unit as an example sentence in the example sentence storage unit (for example, step S901), and the reference language processing result registration unit receives the input sentence together with the input sentence by the input sentence receiving unit. The obtained natural language processing result is registered in the example sentence storage means as the reference language processing result of the input sentence (for example, step S901), and the processing target sentence selection means includes the example sentence stored in the example sentence storage means and the input sentence reception means. Select a processing target sentence from the processing target candidate sentences including the input sentence received by the command and correspond to the selected processing target sentence. Selects a reference language processing result (e.g., step S104 shown in FIG. 12) may be configured so.

  A determination result receiving means (for example, dictionary correction management means 21) for receiving a determination result as to whether or not to register the processing result of the natural language processing for the processing target sentence by the language processing means as the reference language processing result; In response to the determination result being received by the determination result receiving unit being received, the processing target sentence that is the input sentence received by the input sentence receiving unit is registered in the example sentence storage unit as an example sentence, and the reference language processing result registration unit The processing result of the natural language processing by the language processing means for the processing target sentence registered as an example sentence as the reference language processing result of the example sentence in response to the reception of the determination result of registration by the determination result receiving means It may be configured to be registered in the example sentence storage means (for example, step S1001).

  It includes impact analysis abstract creation means (for example, impact analysis abstract creation means 28) for creating an impact analysis abstract based on the analysis result of the dictionary correction impact analysis means, and the dictionary correction impact analysis means includes the contents of the correction instruction indicated by the dictionary editing information. A dictionary correction influence score indicating the influence of dictionary editing according to the process is derived for each sentence to be processed, and the impact analysis abstract creation means arranges the analysis result and the processing result of natural language processing based on the dictionary correction influence score. The analysis abstract may be created, and the analysis result presentation means may be configured to present the impact analysis abstract created by the impact analysis abstract creation means.

  An impact analysis abstract creation means (for example, an impact analysis abstract creation means 28) for creating an impact analysis abstract based on the analysis result of the dictionary correction impact analysis means is provided, and the impact analysis abstract creation means has a plurality of results with the same analysis result. Select one of the sentences to be processed, create an abstract of the impact analysis by arranging the analysis results of the sentences to be processed with different analysis results and the processing results of the natural language processing. An abstract of the impact analysis created by the abstract creation means may be presented.

  An impact analysis abstract creation means (for example, an impact analysis abstract creation means 28) for creating an impact analysis abstract based on the analysis result of the dictionary correction impact analysis means is provided, and the impact analysis abstract creation means includes an analysis result for the processing target sentence. The processing target sentence to be presented is determined based on it, and an abstract of the impact analysis is created by arranging the analysis result of the processing target sentence decided to be the presentation target and the processing result of the natural language processing, and presenting the analysis result The means may be configured to present an impact analysis abstract created by the impact analysis abstract creation means.

  A dictionary registration judging means (for example, dictionary registration judging means 29) for judging whether or not dictionary editing according to the contents of the correction instruction indicated by the dictionary editing information is reflected in the master dictionary based on the analysis result of the dictionary correction influence analyzing means. The analysis result presenting means may be configured to present the determination result of the dictionary registration determining means (see, for example, FIG. 28).

  A dictionary registration determination receiving unit (for example, dictionary correction management unit 21, input device 1) that receives a determination result as to whether or not dictionary editing according to the correction instruction content indicated by the dictionary editing information is reflected in the master dictionary is provided. The means is configured to reflect, in the master dictionary, dictionary editing according to the correction instruction content indicated by the dictionary editing information in accordance with the determination result accepted by the dictionary registration determination accepting means (step S1502). May be.

  The dictionary strengthening support method according to the present invention is a dictionary strengthening support method for supporting a correction work for strengthening a word dictionary in which word information including a word, which is a master dictionary used in natural language processing, is registered, From the step of creating a temporary registration dictionary obtained by editing the word dictionary based on the dictionary editing information indicating the correction instruction content of the word dictionary, and the processing target candidate sentence that is the processing target candidate for the natural language processing, the dictionary editing information Selecting a sentence including a word that is a correction target in step S3 as a processing target sentence, executing a natural language process based on each of the master dictionary and the temporary registration dictionary on the selected processing target sentence, Processing result of natural language processing for processing target sentence based on master dictionary and processing result of natural language processing for processing target sentence based on temporary registration dictionary And analyzing the influence of dictionary editing according to the contents of the correction instruction indicated by the dictionary editing information with respect to a predetermined viewpoint, and presenting the analysis result together with the processing result of the natural language processing It is characterized by.

  With the above-described configuration, it is possible to reduce the workload of dictionary workers when analyzing the effects of dictionary correction based on many differences in analysis results. However, it is possible to assist the dictionary operator so that it can be determined appropriately and easily whether or not dictionary correction should be registered.

  A step of selecting a processing target sentence from processing target candidate sentences including example sentences stored in an example sentence storage unit that stores example sentences prepared in advance as processing target candidate sentences may be executed.

  A step of receiving an input of a natural language processing result of the input sentence together with an input sentence input by a correction operator, and selecting an input sentence including a word to be corrected in the dictionary editing information as a processing target sentence; A step of selecting a natural language processing result corresponding to an input sentence selected as a processing target sentence as a reference language processing result, a processing result of natural language processing for a processing target sentence based on a master dictionary, and a temporary registration dictionary The processing result of the natural language processing for the processing target sentence is compared with the reference language processing result of the selected processing target sentence, and the influence of dictionary editing according to the correction instruction content indicated by the dictionary editing information is determined in advance. The analysis of the current viewpoint and the results of the analysis along with the results of the natural language processing and the reference language processing. It may be configured to perform a flop.

  A step of deriving a dictionary correction influence score indicating the influence of dictionary editing in accordance with the content of the correction instruction indicated by the dictionary editing information for each processing target sentence, and the analysis result and the processing result of the natural language processing are based on the dictionary correction influence score. It may be configured to execute a step of creating an abstract of impact analysis arranged and a step of presenting an abstract of impact analysis.

  Furthermore, the dictionary strengthening support program according to the present invention is a dictionary strengthening support program for supporting a correction work for strengthening a word dictionary in which word information including words, which is a master dictionary used for natural language processing, is registered. A step of creating a temporary registration dictionary obtained by editing the word dictionary based on dictionary editing information indicating the word dictionary correction instruction content on a computer (for example, the dictionary strengthening support system 100, 110, 120, 130, 140, 150) Selecting a sentence including a word that is to be corrected in the dictionary editing information as a processing target sentence from processing target candidate sentences that are candidates for processing of natural language processing, and the selected processing target sentence A natural language process based on each of the master dictionary and the temporary registration dictionary, and the master dictionary. The effect of dictionary editing according to the contents of the correction instruction indicated by the dictionary editing information by comparing the processing result of the natural language processing for the processing target sentence with the processing result of the natural language processing for the processing target sentence based on the temporary registration dictionary Are analyzed with respect to a predetermined viewpoint, and a step of presenting the analysis result together with the processing result of the natural language processing is executed.

  With the above-described configuration, it is possible to reduce the workload of dictionary workers when analyzing the effects of dictionary correction based on many differences in analysis results. However, it is possible to assist the dictionary operator so that it can be determined appropriately and easily whether or not dictionary correction should be registered.

  According to the present invention, the influence of dictionary editing according to the contents of the correction instruction indicated by the dictionary editing information is analyzed from a predetermined viewpoint, and the analysis result is presented together with the processing result of the natural language processing. Can reduce the workload of dictionary workers when analyzing the effects of dictionary corrections based on differences in analysis results, and should dictionary corrections be registered even for dictionary workers who are not skilled in dictionary work? It is possible to assist the dictionary operator so that the determination of whether or not can be made appropriately and easily.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
Embodiment 1 FIG.
FIG. 1 is a block diagram showing an example of the configuration of the dictionary strengthening support system 100 according to the first embodiment of the present invention.

  As shown in FIG. 1, the dictionary strengthening support system 100 includes an input device 1, a data processing device 2, a storage device 3, and an output device 4.

  The input device 1 is a device used for information input such as a keyboard and a mouse. The output device 4 is a device used for information output such as a display device or a printing device.

  The data processing device 2 has a function of executing various processes by program control, for example.

  As shown in FIG. 1, the data processing apparatus 2 includes a dictionary correction management unit 21, a dictionary editing unit 22, a language processing unit 23, a processing target sentence selection unit 24, and a dictionary correction influence analysis unit 25.

  The dictionary correction management means 21 receives the dictionary correction information from the input device 1 and performs a process of outputting the analysis result of the dictionary correction influence of each sentence to the output device 4 together with the language processing result.

  “Dictionary correction information” means information indicating the correction content of the dictionary to be edited. Specifically, for example, word information for the word dictionary to be edited (for example, attribute information indicating grammatical attributes such as words and parts of speech). Information) including the registration, deletion, and change contents.

  The dictionary editing unit 22 performs a process of editing a dictionary to be edited based on a dictionary editing request from the dictionary correction management unit 21 and storing the edited result in the temporary registration dictionary storage unit 32 as a temporary registration dictionary.

  The language processing unit 23 performs language processing of the processing target sentence based on a word dictionary (master dictionary, temporary registration dictionary), and performs processing for outputting a language processing result.

  The processing target sentence selection unit 24 selects a sentence including a word (for example, a word) to be edited by the dictionary editing unit 22 from among input sentences input as processing target candidates for the language processing by the input device 1. Perform processing to select as a sentence.

  The dictionary correction influence analysis means 25 receives the language processing result of the processing target sentence based on each of the master dictionary and the temporary registration dictionary together with the processing target sentence, and performs the influence analysis registered in the influence analysis table storage unit 33 in advance. Therefore, it has various functions such as a function for executing a comparative analysis of language processing results for each processing target sentence based on various viewpoints.

  The storage device 3 is configured by a storage medium that stores various types of information such as a database device, for example, and as illustrated in FIG. 1, a master dictionary storage unit 31, a temporary registration dictionary storage unit 32, and an impact analysis table storage unit 33. Including.

  The master dictionary storage unit 31 is a storage medium that stores a master dictionary in which word information is described in a data format readable by the language processing unit 23.

  The temporary registration dictionary storage unit 32 is a storage medium that stores a temporary registration dictionary that is a result of dictionary editing of the master dictionary in a data format that can be read by the language processing unit 23.

  The impact analysis table storage unit 33 is a storage medium that stores an impact analysis table. The impact analysis table is pre-registered by associating the viewpoint of comparative analysis for determining the impact of dictionary correction from the language processing results based on the master dictionary and the temporary registration dictionary, and data indicating the estimated impact. (See FIG. 21).

  Next, the operation of the dictionary strengthening support system 100 in the first embodiment of the present invention will be described. FIG. 2 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 100.

  First, the dictionary operator operates the input device 1 to input dictionary correction information for instructing registration, deletion, and change of word information for the master dictionary. In response to an input operation by the dictionary operator, the input device 1 transmits dictionary correction information to the data processing device 2.

  The dictionary correction management means 21 receives the dictionary correction information from the input device 1 (step S101), and in order to request the creation of a temporary registration dictionary reflecting the dictionary correction indicated by the received dictionary correction information, the dictionary editing means 22 A dictionary editing request is sent to.

  When the dictionary editing request is received, the dictionary editing unit 22 generates a temporary registration dictionary in which the dictionary correction indicated by the dictionary correction information is performed on the master dictionary stored in the master dictionary storage unit 31, and the temporary registration dictionary storage unit 32 (step S102). In step S102, the master dictionary stored in the master dictionary storage unit 31 is maintained as it is without being edited.

  When the temporary registration dictionary is stored in the temporary registration dictionary storage unit 32 (for example, when there is a temporary registration dictionary in which a plurality of dictionary corrections and some dictionary corrections are reflected), the dictionary editing unit 22 In step S102, dictionary correction indicated by the dictionary correction information is performed on the temporary registration dictionary stored in the temporary registration dictionary storage unit 32, and the temporary registration dictionary stored in the temporary registration dictionary storage unit 32 is updated.

  Next, the dictionary operator operates the input device 1 to input a sentence (language processing target candidate sentence) as a candidate for a language processing target.

  In response to an input operation by the dictionary worker, the input device 1 transmits a sentence (input sentence) input by the dictionary worker to the data processing device 2.

  The dictionary correction management unit 21 receives the input sentence from the input device 1 (step S103), and transmits a processing target sentence selection request together with the received input sentence to the processing target sentence selecting unit 24.

  When the processing target sentence selection request is received, the processing target sentence selecting means 24 selects a sentence including the surface layer of the dictionary correction target word from the received input sentences as the processing target sentence (step S104).

  The “sentence including the surface layer of the dictionary correction target word” means a sentence including the word that is the target of dictionary correction. That is, the “surface layer of the target word” is the word itself that is the target of dictionary correction, and indicates a portion excluding attribute information.

  When the processing target sentence is selected, the processing target sentence selection unit 24 transmits the selected processing target sentence to the dictionary correction management unit 21.

  When the processing target sentence is received, the dictionary correction management unit 21 transmits a request for creating a language processing result of the processing target sentence based on each of the master dictionary and the temporary registration dictionary to the language processing unit 23.

  When the language processing result creation request is received, the language processing unit 23 executes the language processing of the processing target sentence based on each of the master dictionary and the temporary registration dictionary (step S105), and sends the language processing result to the dictionary correction management unit 21. Send to.

  When the language processing result is received, the dictionary correction management unit 21 transmits the received language processing result and the processing target sentence to the dictionary correction influence analyzing unit 25.

  Upon receiving the language processing result and the processing target sentence, the dictionary correction influence analyzing unit 25 analyzes the language processing result based on the master dictionary and the language processing result based on the temporary registration dictionary in order to analyze the influence of the dictionary correction. The impact analysis table stored in the impact analysis table storage unit 33 is used for comparative analysis from various viewpoints (step S106).

  The viewpoints used for analysis include, for example, “generation and disappearance of unregistered words by correction”, “decrease and increase of morpheme numbers”, “generation of word-cutting boundaries that are not in the envelope-inclusive relationship”, and “generation of words to be corrected” And “disappearance”, “generation and disappearance of function words”, “no difference”, and the like. In step S106, the difference between the language processing results before and after the dictionary correction is analyzed from the above viewpoints.

  When the analysis result of the language processing result is obtained, the dictionary correction influence analyzing unit 25 transmits the analysis result to the dictionary correction managing unit 21.

  The dictionary correction management unit 21 transmits the received analysis result and the language processing result to the output device 4. The output device 4 outputs the received analysis results and language processing results, and presents them to the dictionary operator (step S107).

  The dictionary operator confirms the presented analysis result and the language processing result, and the effect of the dictionary correction on the language processing result is improved (meaning that the language processing result is improved by the dictionary correction). Or a side effect (meaning that the dictionary processing has an adverse effect on the language processing result).

  Then, the dictionary operator may reflect the dictionary correction in the master dictionary if it is determined to be an improvement, and may not reflect the dictionary correction in the master dictionary if it is determined as a side effect.

  As described above, in the first embodiment, dictionary correction influence analysis means for performing comparative analysis on the language processing results obtained based on the master dictionary and the temporary registration dictionary based on various viewpoints. 25, and the analysis result is presented together with the language processing result, so that the dictionary operator can easily grasp the influence of the dictionary correction on the language processing result, and the dictionary correction should be reflected in the master dictionary. It is possible to assist the dictionary operator so that the determination of whether or not can be performed appropriately and easily.

  For this reason, even dictionary workers who are not familiar with language processing systems and are not skilled in dictionary work will be able to analyze the impact of dictionary correction and dictionary correction based on the difference in a large amount of analysis results (language processing results). Registration decision can be made appropriately and easily. In addition, it is possible to reduce the workload of the dictionary operator when analyzing the influence of dictionary correction based on many analysis result differences.

Embodiment 2. FIG.
Next, a second embodiment of the present invention will be described.
FIG. 3 is a block diagram showing an example of the configuration of the dictionary strengthening support system 110 according to the second embodiment of the present invention. In FIG. 3, parts having the same configuration as the dictionary strengthening support system 100 shown in FIG. 1 described above are assigned the same reference numerals and detailed description thereof is omitted.

  As shown in FIG. 3, the dictionary strengthening support system 110 includes an input device 1, a data processing device 2, a storage device 3 </ b> A, and an output device 4.

  The storage device 3A is configured by a storage medium that stores various types of information such as a database device, for example, and as illustrated in FIG. 3, a master dictionary storage unit 31, a temporary registration dictionary storage unit 32, and an impact analysis table storage unit 33. And an example sentence storage unit 34.

  That is, the dictionary strengthening support system 110 is configured such that the storage device 3 further includes the example sentence storage unit 34 as compared with the dictionary strengthening support system 100 described above.

  The example sentence storage unit 34 is a storage medium that stores example sentences prepared in advance. In this example, the processing target sentence selection unit 24 has a function of selecting, as a processing target sentence, an example sentence that includes a surface layer of a dictionary correction target word stored in the example sentence storage unit 34.

  Next, the operation of the dictionary strengthening support system 110 in the second embodiment of the present invention will be described. FIG. 4 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 110.

  In this example, in the dictionary strengthening support process, the dictionary strengthening support system 110 performs the processes of steps S101 to S102 described above. Next, the dictionary correction management unit 21 transmits a processing target sentence selection request to the processing target sentence selection unit 24 without performing step S103.

  When the processing target sentence selection request is received, the processing target sentence selecting unit 24 includes the surface layer of the target word for dictionary correction from the example sentences (used as language processing target candidate sentences) stored in the example sentence storage unit 34. The sentence is selected as a processing target sentence (step S201), and the selected processing target sentence is transmitted to the dictionary correction management means 21.

  Thereafter, the dictionary strengthening support system 110 performs the processes of steps S105 to S107 described above.

  As described above, in the second embodiment, since the example sentence storage unit 34 for registering language processing target candidate sentences in advance is provided, the processing target sentence is obtained from a large number of pre-registered example sentences. It becomes possible to select and use. Therefore, it is possible to eliminate the necessity of inputting a language processing target sentence each time a dictionary is corrected, and to easily perform an influence analysis of dictionary correction using various contexts as processing target sentences.

Embodiment 3 FIG.
Next, a third embodiment of the present invention will be described.
FIG. 5 is a block diagram showing an example of the configuration of the dictionary strengthening support system 120 according to the third embodiment of the present invention. In FIG. 5, parts having the same configuration as the dictionary strengthening support system 100 shown in FIG. 1 and the dictionary strengthening support system 110 shown in FIG. Is omitted.

  As shown in FIG. 5, the dictionary strengthening support system 120 includes an input device 1, a data processing device 2 </ b> A, a storage device 3 </ b> A, and an output device 4.

  The data processing apparatus 2A includes a dictionary correction management unit 21, a dictionary editing unit 22, a language processing unit 23, a processing target sentence selection unit 24, a dictionary correction influence analysis unit 25, and an example sentence registration unit 26.

  That is, the dictionary strengthening support system 120 is configured such that the data processing device 2 further includes example sentence registration means 26 as compared with the dictionary strengthening support system 110 described above.

  The example sentence registration unit 26 has a function of performing processing for storing the input sentence input from the input device 1 by the operation of the dictionary operator as an example sentence in the example sentence storage unit 34 in accordance with the example sentence registration request from the dictionary correction management unit 21. .

  That is, in this example, the dictionary correction management means 21 receives an example sentence registration request for an input sentence input from the input device 1 by the operation of the dictionary operator, and transmits an example sentence registration request for the input sentence to the example sentence registration means 26. Process.

  Next, operation | movement of the dictionary reinforcement | strengthening assistance system 120 in the 3rd Embodiment of this invention is demonstrated. FIG. 6 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 120.

  In this example, in the dictionary strengthening support process, in the dictionary strengthening support system 120, after performing the above-described steps S101 to S103, the dictionary correction management means 21 registers the received input sentence and example sentence registration request as an example sentence. Processing to transmit to the means 26 is performed.

  When the example sentence registration request is received, the example sentence registration unit 26 performs a process of storing the received input sentence as an example sentence in the example sentence storage unit 34 (step S301).

  Next, the dictionary correction management unit 21 transmits a processing target sentence selection request to the processing target sentence selection unit 24.

  Thereafter, the dictionary strengthening support system 110 performs the above-described processing of step S201, step S105 to step S107.

  As described above, in the third embodiment, the example sentence registration unit 26 for additionally registering example sentences in the example sentence storage unit 34 is provided. Even if it is not in the example sentence registered in the storage unit 34, the example sentence registered in the example sentence storage unit 34 can be additionally registered.

  In this way, even when a sentence that can be selected as a processing target sentence does not exist in the example sentence at the time of dictionary correction, by accepting an input of a sentence including the surface layer of the target word for dictionary correction, Impact analysis can be performed. Further, it is possible to easily analyze the influence of dictionary correction in various contexts without inputting a sentence to be processed for language processing each time the dictionary is corrected.

  That is, for example, when adding a new word by dictionary correction, there is no guarantee that an example sentence prepared in advance includes an example sentence including the word to be corrected. According to the third embodiment described above, a framework for expanding example sentences including words to be registered when the word dictionary is to be strengthened is presented. The example sentences can be expanded simultaneously to cover them.

Embodiment 4 FIG.
Next, a fourth embodiment of the present invention will be described.
The dictionary strengthening support system in the fourth embodiment has the same configuration as the dictionary strengthening support system 110 shown in FIG. 3, and will be described with reference to the dictionary strengthening support system 110 in FIG. Detailed description thereof is omitted.

  In the present embodiment, the processing target sentence selection unit 24 instructs the dictionary correction management unit 21 when there is no sentence including the surface layer of the dictionary correction target word in the example sentence registered in the input sentence and example sentence storage unit 34. On the other hand, a process of notifying that there is no sentence that can be selected as a process target sentence (notifying that there is no process target sentence) is performed.

  Further, the dictionary correction management means 21, in response to receiving a notification that there is no sentence that can be selected as a processing target sentence, outputs an input request to the dictionary operator for a sentence to be subjected to language processing. 4 is executed.

  Next, the operation of the dictionary strengthening support system 110 according to the fourth embodiment of the present invention will be described. FIG. 7 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 110.

  In this example, in the dictionary strengthening support process, the dictionary strengthening support system 110 performs the processing in steps S101 to S102 described above, and then sends a processing target sentence selection request by the dictionary correction management means 21 to the processing target sentence selecting means 24. Send to.

  When the processing target sentence selection request is received, the processing target sentence selecting unit 24 selects, from the example sentences stored in the example sentence storage unit 34, a sentence including the surface layer of the target word for dictionary correction as a processing target sentence (step S1). S201).

  If a processing target sentence has been selected in step S201 (Y in step S401), the processing target sentence selection unit 24 transmits the selected processing target sentence to the dictionary correction management unit 21. Then, step S105-step S107 mentioned above are performed.

  On the other hand, in the example sentence stored in the example sentence storage unit 34, there is no sentence including the surface layer of the target word of the dictionary correction, and no processing target sentence has been selected in step S201 (N in step S401). ), The processing target sentence selection unit 24 notifies the dictionary correction management unit 21 that there is no processing target sentence among the selection candidate sentences.

  Upon receiving notification that the processing target sentence does not exist, the dictionary correction management means 21 requests the output device 4 to input the processing target sentence in order to instruct the dictionary operator to input the processing target sentence. An input request instruction for presenting is made (step S402).

  In response to the input request instruction, the output device 4 presents information for prompting input of a sentence to be subjected to language processing. For example, the output device 4 performs a process of displaying on the screen a word indicating a request to input a processing target sentence to the dictionary worker.

  When a sentence is input from the input device 1 in response to the input request, the dictionary correction management unit 21 accepts the input sentence (Y in step S403), and processes the input sentence and a request for selecting a processing target sentence. It transmits to the sentence selection means 24. When receiving the request for selecting the processing target sentence, the processing target sentence selecting unit 24 selects the processing target sentence from the input sentences (step S104). Then, the process proceeds to step S401.

  If no input is made even after a predetermined time has elapsed since the input request instruction was issued in step S402, and the input sentence was not accepted by the dictionary correction management means 21 (N in step S403). ), The process is terminated.

  As described above, in the fourth embodiment, when there is no sentence to be selected as a processing target sentence in pre-registered example sentences, an instruction to input a sentence to be processed in language processing is given. Because it is structured, even if there is no sentence to be processed in the pre-registered example sentence, the dictionary operator can be prompted to input an appropriate sentence that includes the word to be corrected in the surface layer. The selected sentence is selected as a processing target sentence, and the influence of dictionary correction can be analyzed.

  That is, as a result of dictionary correction, when an example sentence including a word that is the target of dictionary correction is not registered in advance, input of a sentence used as a processing target sentence for analyzing the influence of dictionary correction is accepted and accepted. Since the sentence is added to the example sentence, the example sentence can be expanded so that the influence of the dictionary correction can be appropriately analyzed for the unregistered word added in the dictionary strengthening.

Embodiment 5 FIG.
Next, a fifth embodiment of the present invention will be described.
Since the dictionary strengthening support system in the fifth embodiment has the same configuration as the dictionary strengthening support system 120 shown in FIG. 5 described above, the dictionary strengthening support system 120 will be described with reference to the dictionary strengthening support system 120 in FIG. Detailed description thereof is omitted.

  In this embodiment, the processing target sentence selection unit 24 indicates whether each processing target sentence is selected from an input sentence or an example sentence when the processing target sentence is selected and transmitted to the dictionary correction management unit 21. It has a function of adding information to each processing target sentence.

  The sentence selected from the input sentence as the processing target sentence is referred to as the input sentence origin, and the sentence selected as the processing target sentence from the example sentence is referred to as the example sentence origin. Accordingly, the origin information is information indicating whether the input sentence originates or the example sentence originates.

  Further, in this embodiment, the dictionary correction management means 21 determines that all of the processing target sentences originate from the input sentence and have no example sentence origin based on the processing target sentence and the origin information from the processing target sentence selection means 24. When the determination is made, an example sentence registration request is transmitted to instruct the example sentence registration means 26 to register the input sentence as an example sentence.

  Next, the operation of the dictionary strengthening support system 120 in the fifth embodiment of the present invention will be described. FIG. 8 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 120.

  In this example, in the dictionary strengthening support process, the dictionary strengthening support system 120 performs the processes of steps S101 to S103 described above.

  Next, when the processing target sentence selection unit 24 receives the processing target sentence selection request from the dictionary correction management unit 21, the processing target sentence selection unit 24 selects the processing target sentence from the input sentence and the example sentence stored in the example sentence storage unit 34. (Step S104, Step S201).

  When the processing target sentence is selected, the processing target sentence selecting unit 24 adds the origin information to each of the selected processing target sentences, and transmits the processing target sentence to which the origin information is added to the dictionary correction management unit 21.

  The dictionary correction management unit 21 confirms the origin information added to the processing target sentence received from the processing target sentence selection unit 24 and confirms whether or not the received processing target sentence has an example sentence origin (step S501). ).

  If there is a sentence to be processed originating in an example sentence (Y in step S501), the dictionary strengthening support system 120 performs the processes in steps S105 to S107 described above.

  If there is no sentence to be processed originating in the example sentence (N in step S501), the dictionary correction management means 21 transmits the input sentence accepted in step S103 to the example sentence registration means 26, and uses the input sentence as an example sentence. An example sentence registration request for registration is transmitted (step S502).

  When the example sentence registration request is received, the example sentence registration unit 26 performs a process of registering the input sentence accepted in step S103 as an example sentence in the example sentence storage unit 34 (step S301). Thereafter, the processes of steps S105 to S107 described above are performed.

  As described above, in the fifth embodiment, when it is determined whether or not there are no processing target sentences originating from example sentences and all processing target sentences originated from input sentences, and there is no processing target sentence originating from example sentences Since it is configured to instruct the input sentence to be registered, if the sentence including the dictionary correction target word is not registered as an example sentence, the input sentence including the dictionary correction target word can be registered as an example sentence, Example sentences can be enhanced to cover the target words of dictionary correction. Therefore, the completeness of the example sentence can be improved in parallel with the dictionary correction.

Embodiment 6 FIG.
Next, a sixth embodiment of the present invention will be described.
Since the dictionary strengthening support system in the sixth embodiment has the same configuration as the dictionary strengthening support system 120 shown in FIG. 5 described above, the dictionary strengthening support system 120 will be described with reference to the dictionary strengthening support system 120 in FIG. Detailed description thereof is omitted.

  In this embodiment, as in the fifth embodiment described above, the processing target sentence selection unit 24 inputs each processing target sentence when the processing target sentence is selected and transmitted to the dictionary correction management unit 21. It has a function of adding origin information indicating whether a sentence or an example sentence is selected to each sentence to be processed.

  Further, in this embodiment, the dictionary correction management means 21 is based on the processing target sentence and origin information from the processing target sentence selection means 24 and the analysis result from the dictionary correction influence analysis means 25, and is processed from the input sentence origin. In order to instruct the example sentence registration means 26 to register the input sentence as an example sentence when it is determined that the analysis result obtained for the sentence is different from any analysis result for the sentence to be processed originating in the example sentence. Make a registration request.

  Next, the operation of the dictionary strengthening support system 120 in the sixth embodiment of the present invention will be described. FIG. 9 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 120.

  As shown in FIG. 9, in this example, in the dictionary strengthening support process, the dictionary strengthening support system 120 performs the processes of steps S101 to S103 described above.

  Next, when the processing target sentence selection unit 24 receives the processing target sentence selection request from the dictionary correction management unit 21, the processing target sentence selection unit 24 selects the processing target sentence from the input sentence and the example sentence stored in the example sentence storage unit 34. (Step S104, Step S201).

  When the processing target sentence is selected, the processing target sentence selecting unit 24 adds the origin information to each of the selected processing target sentences, and transmits the processing target sentence to which the origin information is added to the dictionary correction management unit 21. Thereafter, the dictionary strengthening support system 120 executes Step S105 and Step S106.

  After the influence analysis of the dictionary correction is performed, the dictionary correction management means 21 inputs the input sentence based on the processing target sentence and origin information from the processing target sentence selection means 24 and the analysis result from the dictionary correction influence analysis means 25. It is confirmed whether or not the analysis result obtained for the processing sentence with the origin matches any of the analysis results with respect to the processing sentence with the example sentence origin (step S601).

  If it matches any of the analysis results for the sentence to be processed originating in the example sentence (N in step S601), the dictionary strengthening support system 120 performs the process in step S107 described above.

  On the other hand, if it does not coincide with any of the analysis results for the sentence subject to processing of the example sentence (Y in step S601), the dictionary correction management means 21 analyzes the sentence subject to processing of the example sentence for the example sentence registration means 26. An input sentence that is an analysis result that does not match any of the results is transmitted, and an example sentence registration request is made to register the input sentence as an example sentence (step S502).

  When the example sentence registration request is received, the example sentence registering means 26 outputs an input sentence that becomes an analysis result that does not match any of the analysis results for the sentence to be processed among the input sentences received in step S103, to the example sentence storage unit. A process of registering as an example sentence in 34 is performed (step S301). Then, the process of step S107 mentioned above is performed.

  As described above, in the sixth embodiment, it is confirmed whether or not the analysis result obtained for the sentence to be processed originating from the input sentence matches one of the analysis results for the sentence to be processed originating in the example sentence. When it is determined that the analysis result is different from any analysis result, it is configured to instruct the input sentence that has a different analysis result from the analysis result for the sentence to be processed that originates in the example sentence. It is possible to improve the completeness of the context in the example sentences necessary for doing so. Further, with the above configuration, it is possible to improve the completeness of example sentences in parallel with dictionary correction.

Embodiment 7 FIG.
Next, a seventh embodiment of the present invention will be described.
Since the dictionary strengthening support system in the seventh embodiment has the same configuration as the dictionary strengthening support system 100 shown in FIG. 1 described above, it will be described with reference to the dictionary strengthening support system 100 in FIG. Detailed description thereof is omitted.

  In the present embodiment, the input device 1 has a function of inputting the language processing result together with the sentence based on the operation of the dictionary worker.

  Further, when the language processing result is given together with the input sentence, the processing target sentence selecting unit 24 selects the input sentence as the processing target sentence, and uses the language processing result as the reference language processing result for the processing target sentence. To select both.

  Further, the dictionary correction influence analysis means 25 not only determines the influence of the dictionary correction from the language processing results based on the master dictionary and the temporary registration dictionary based on the influence analysis table, but also includes the master dictionary and the temporary registration dictionary. It also has a function of determining the influence of dictionary correction by comparing each language processing result with the reference language processing result.

  Further, the dictionary correction management means 21 performs a process of presenting the analysis result together with the language processing result and the reference language processing result from the output device 4.

  Next, the operation of the dictionary strengthening support system 100 in the seventh embodiment of the present invention will be described. FIG. 10 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 100.

  In this example, in the dictionary strengthening support process, the dictionary strengthening support system 100 performs the processes of steps S101 to S102 described above.

  Next, in this example, the dictionary operator operates the input device 1 to input a sentence as a candidate for a language processing target and a language processing result for the sentence.

  In response to the input operation by the dictionary operator, the input device 1 transmits the language processing target sentence and the language processing result to the data processing device 2.

  The dictionary correction management unit 21 receives the input sentence and the language processing result from the input device 1 (step S701), and transmits the received input sentence and the language processing result to the processing target sentence selecting unit 24 for processing. Request selection of the target sentence.

  Upon receiving the processing target sentence selection request, the processing target sentence selecting unit 24 selects, from the received language processing target candidate sentences, a sentence including the surface layer of the dictionary correction target word as a processing target sentence, and receives the received language. The processing result is selected as the reference language processing result (step S104).

  When the processing target sentence and the reference language processing result are selected, the processing target sentence selecting unit 24 transmits the selected processing target sentence and the reference language processing result to the dictionary correction management unit 21.

  When the processing target sentence is received, the dictionary correction management unit 21 requests the language processing unit 23 to create a language processing result of the processing target sentence based on each of the master dictionary and the temporary registration dictionary.

  When the language processing result creation request is received, the language processing unit 23 executes the language processing (step S105), and transmits the language processing result to the dictionary correction management unit 21.

  When the language processing result is received, the dictionary correction management unit 21 transmits the received language processing result, the processing target sentence, and the reference language processing result to the dictionary correction influence analyzing unit 25.

  When the language processing result, the processing target sentence, and the reference language processing result are received, the dictionary correction influence analyzing unit 25 converts the language processing result based on the master dictionary and the language processing result based on the temporary registration dictionary into the influence analysis table storage unit 33. Are analyzed from various viewpoints using the impact analysis table stored in the table (step S106).

  In this example, in step S106, the dictionary correction influence analysis means 25 performs the dictionary correction influence analysis by comparing the language processing results of the master dictionary and the temporary registration dictionary with the reference language processing results.

  When the analysis result of the language processing result is obtained, the dictionary correction influence analyzing unit 25 transmits the analysis result to the dictionary correction managing unit 21.

  The dictionary correction management means 21 transmits the received analysis result, language processing result, and reference language processing result to the output device 4. The output device 4 performs processing for presenting the received analysis result, language processing result, and reference language processing result to the dictionary operator (step S107).

  As described above, in the seventh embodiment, not only the influence of dictionary correction is determined from the language processing results based on the master dictionary and the temporary registration dictionary based on the influence analysis table, but also the master dictionary and Since it is configured to determine the influence of dictionary correction by comparing each language processing result of the temporary registration dictionary and the reference language processing result, the influence of the dictionary correction on the analysis result is evaluated based on the reference language processing result. be able to.

  In addition, since it is configured to determine the influence of dictionary correction by comparing the language processing result of each of the master dictionary and the temporary registration dictionary with the reference language processing result, even if it is not preferable that the analysis result changes. Since the reference analysis result is set as the reference language processing result, it is possible to easily determine whether the change in the analysis result of the example sentence before and after the dictionary correction is within an allowable range. Therefore, even when dictionary strengthening is continued for a long time or performed by a plurality of dictionary workers, consistency and consistency in the entire dictionary work can be easily maintained.

  Specifically, for example, when correcting a previously corrected word again, if the reference language processing result assumed in the previous correction is used, consistency and consistency in the entire dictionary work can be maintained. It becomes easy.

Embodiment 8 FIG.
Next, an eighth embodiment of the present invention will be described.
The dictionary strengthening support system in the eighth embodiment has the same configuration as the dictionary strengthening support system 110 shown in FIG. 3, and will be described with reference to the dictionary strengthening support system 110 in FIG. Detailed description thereof is omitted.

  As shown in FIG. 3, the dictionary strengthening support system 110 according to the present embodiment is a storage device 3 of the dictionary strengthening support system 100 that executes the dictionary strengthening support processing (see FIG. 10) in the seventh embodiment described above. In addition, an example sentence storage unit 34 is added.

  In this embodiment, the example sentence storage unit 34 stores an example sentence prepared in advance as a sentence to be processed by the language processing unit 23 together with a reference language processing result for the sentence.

  In addition, the processing target sentence selection unit 24 executes a process of selecting a corresponding reference language processing result when selecting a processing target sentence from the example sentences registered in the example sentence storage unit 34.

  Next, dictionary strengthening support processing of the dictionary strengthening support system 110 according to the eighth embodiment of the present invention will be described with reference to FIG.

  As shown in FIG. 4, the dictionary strengthening support process in the present embodiment excludes input sentences and the like (step S <b> 701) from the process of the seventh embodiment (see FIG. 10) described above, and The selection target is not an input sentence (step S104) but an example sentence (step S201) registered in the example sentence storage unit 34.

  In the dictionary strengthening support process in the present embodiment, in step S201 shown in FIG. 4, the processing target sentence selection unit 24 processes a sentence including the word to be corrected from the example sentence stored in the example sentence storage unit 34 on the surface layer. The target sentence is selected and transmitted to the dictionary correction management means 21 together with the reference language processing result of the processing target sentence. Thereafter, the same processing as the processing of the seventh embodiment (see FIG. 10) described above is performed.

  As described above, in the eighth embodiment, the configuration includes the example sentence storage unit 34 that stores a large number of example sentences together with the reference language processing result in advance. Even if the reference language processing result is not input, it is possible to obtain the effect of the seventh embodiment described above, and to easily perform the influence analysis of dictionary correction in various contexts.

Embodiment 9 FIG.
Next, a ninth embodiment of the present invention will be described.
FIG. 11 is a block diagram showing an example of the configuration of the dictionary strengthening support system 130 according to the ninth embodiment of the present invention. In FIG. 11, parts having the same configuration as that of the dictionary strengthening support system 120 shown in FIG. 5 described above are assigned the same reference numerals and detailed description thereof is omitted.

  As shown in FIG. 11, the dictionary strengthening support system 130 includes an input device 1, a data processing device 2 </ b> B, a storage device 3 </ b> A, and an output device 4.

  The data processing device 2B includes a dictionary correction management unit 21, a dictionary editing unit 22, a language processing unit 23, a processing target sentence selection unit 24, a dictionary correction influence analysis unit 25, an example sentence registration unit 26, and a reference language processing. Result registration means 27.

  That is, in the dictionary strengthening support system 130, the data processing device 2 </ b> A further includes the reference language processing result registration unit 27 as compared with the dictionary strengthening support system 120 described above.

  The reference language processing result registration unit 27 has a function of performing processing such as storing the input sentence input by the input device 1 and the reference language processing result in the example sentence storage unit 34 based on the operation of the dictionary operator.

  That is, in this example, the dictionary correction management means 21 receives an example sentence registration request for an input sentence input from the input device 1 by the operation of the dictionary operator, and transmits an example sentence registration request for the input sentence to the example sentence registration means 26. At the same time, processing for transmitting a registration request for the reference language processing result of the input sentence to the reference language processing result registration means 27 is performed.

  Next, the operation of the dictionary strengthening support system 130 in the ninth embodiment of the present invention will be described. FIG. 12 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 130.

  As shown in FIG. 12, in this example, in the dictionary strengthening support process, the dictionary strengthening support system 130 performs the processes of steps S101 to S102 described above.

  Next, in this example, the dictionary operator operates the input device 1 to input a sentence as a candidate for a language processing target and a language processing result for the sentence.

  In response to the input operation by the dictionary operator, the input device 1 transmits the language processing target sentence and the language processing result to the data processing device 2B.

  The dictionary correction management unit 21 receives the language processing target candidate sentence and the language processing result from the input device 1 (step S701), transmits a registration request as an example sentence of the input sentence to the example sentence registration unit 26, and includes the input sentence. A registration request for registering the input language processing result as a reference language processing result is transmitted to the reference language processing result registration means 27.

  In response to the registration request, the example sentence registration unit 26 registers the input sentence as an example sentence in the example sentence storage unit 34 (step S901). In response to the registration request, the reference language processing result registration unit 27 registers the language processing result input for the input sentence in the example sentence storage unit 34 as the reference language processing result for the input sentence (step S901).

  Thereafter, similarly to the above-described seventh embodiment, the processes of steps S104 to S107 are executed.

  As described above, in the ninth embodiment, by providing the reference language processing result registration means 27, not only a large number of example sentences are stored in advance together with the reference language processing result, but also the target of dictionary correction. Even if there is no processing target sentence for the selected word in the example sentence, the example sentence and its reference language processing result can be additionally registered. Accordingly, it is possible to easily analyze the influence of dictionary correction in various contexts without inputting the language processing target sentence and its reference language processing result every time the dictionary is corrected.

Embodiment 10 FIG.
Next, a tenth embodiment of the present invention will be described.
Since the dictionary strengthening support system in the tenth embodiment has the same configuration as the dictionary strengthening support system 130 shown in FIG. 11 described above, it will be described with reference to the dictionary strengthening support system 130 in FIG. Detailed description thereof is omitted.

  In the present embodiment, the dictionary correction management means 21 receives a registration request for registering a language processing result for a processing target sentence as a reference language processing result from the input device 1 and registers an example sentence registration request for registering the processing target sentence as an example sentence. Is sent to the example sentence registration unit 26, and a registration request for registering the language processing result of the example sentence selected as the processing target sentence as the reference language processing result for the example sentence is sent to the reference language processing result registration unit 27. Process to send.

  Next, the operation of the dictionary strengthening support system 130 in the tenth embodiment of the present invention will be described. FIG. 13 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 130.

  In this example, in the dictionary strengthening support process, the dictionary strengthening support system 130 performs the processes of steps S101 to S103, steps S104 and S201, and steps S105 to S107 described above.

  In this example, after the language processing result and the analysis result for the sentence to be processed are presented from the output device 4 in step S107, the process waits for an input request for a reference language processing result from the input device 1.

  When a registration request as a reference language processing result is input from the input device 1 in the waiting state, the dictionary correction management means 21 responds to the registration request when the processing target sentence is already registered in the example sentence. Transmits a registration request for registering the language processing result as the reference language processing result for the example sentence to the reference language processing result registration means 27.

  In addition, if the processing target sentence is not registered in the example sentence, the dictionary correction management unit 21 transmits a registration request as an example sentence of the processing target sentence to the example sentence registration unit 26, and then the reference language processing result registration unit 27. Send a registration request to.

  When the reference language processing result registration unit 27 receives a language processing result registration request from the dictionary correction management unit 21, the language processing result obtained for the processing target sentence in step S105 is registered as an example sentence. The result is registered in the example sentence storage unit 34 as a reference language processing result of the processing target sentence (step S1001).

  As described above, in the tenth embodiment, a language processing result that can be regarded as a correct answer or a language processing result that is not preferably changed in dictionary correction is selected from the language processing results obtained for the processing target sentence. And the reference language processing result for the example sentence can be easily accumulated.

Embodiment 11 FIG.
Next, an eleventh embodiment of the present invention will be described.
FIG. 14 is a block diagram showing an example of the configuration of the dictionary strengthening support system 140 in the eleventh embodiment of the present invention. In FIG. 14, parts having the same configuration as that of the dictionary strengthening support system 100 shown in FIG. 1 described above are assigned the same reference numerals and detailed description thereof is omitted.

  As shown in FIG. 14, the dictionary strengthening support system 140 includes an input device 1, a data processing device 2 </ b> C, a storage device 3, and an output device 4.

  The data processing device 2C includes a dictionary correction management unit 21, a dictionary editing unit 22, a language processing unit 23, a processing target sentence selection unit 24, a dictionary correction influence analysis unit 25, and an influence analysis abstract creation unit 28. .

  That is, the dictionary strengthening support system 140 is configured such that the data processing apparatus 2 further includes an influence analysis abstract creating means 28 as compared with the dictionary strengthening support system 100 described above.

  The impact analysis abstract creation means 28 calculates the dictionary correction impact score based on the impact analysis result for each processing target sentence from the dictionary correction impact analysis means 25, and the effect analysis result of the processing target sentence in the order of the value of the dictionary correction impact score. It has a function to create an abstract of impact analysis by rearranging the results of language processing.

  The dictionary correction management means 21 performs a process of transmitting the impact analysis abstract created by the impact analysis abstract creation means 28 to the output device 4.

  Next, the operation of the dictionary strengthening support system 140 in the eleventh embodiment of the present invention will be described. FIG. 15 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 140.

  As shown in FIG. 15, in this example, in the dictionary strengthening support process, the dictionary strengthening support system 140 performs the processes of steps S101 to S106 described above.

  In the present embodiment, in step S106, the dictionary modification impact analysis unit 25 stores the language processing result based on the master dictionary and the language processing result based on the temporary registration dictionary in the impact analysis table storage unit 33. The analysis is performed according to the impact analysis table and the analysis result is output.

  Further, the dictionary correction influence analysis means 25 calculates a dictionary correction influence score for each analysis result, and transmits it to the impact analysis abstract creation means 28.

  Upon receiving the analysis result from the dictionary correction impact analysis means 25, the impact analysis abstract creation means 28 creates an impact analysis result abstract arranged in the order of the dictionary correction impact score values (step S1101), and this is dictionary corrected. Output to the management means 21.

  The dictionary correction management means 21 transmits the abstract of the impact analysis result to the output device 4, and causes the output device 4 to present the abstract of the impact analysis result (step S1102).

  As explained above, the analysis results for each processing target sentence are arranged in the order of the dictionary correction impact score values, so the analysis result is presented. A high processing target sentence and a processing target sentence that is highly likely to have side effects can be presented separately, and the dictionary operator examines the analysis results of a large amount of processing target sentences and determines dictionary registration In this case, it becomes possible to quickly grasp the trend of the analysis result. For this reason, even a dictionary worker who is not familiar with the language processing system and is not skilled in dictionary work can easily analyze the influence of dictionary correction on a large amount of analysis result differences and determine dictionary correction registration. Can do.

Embodiment 12 FIG.
Next, a twelfth embodiment of the present invention will be described.
Since the dictionary strengthening support system in the twelfth embodiment has the same configuration as the dictionary strengthening support system 140 shown in FIG. 14 described above, the dictionary strengthening support system 140 will be described with reference to the dictionary strengthening support system 140 in FIG. Detailed description thereof is omitted.

  In the present embodiment, the abstract generation means 28 for impact analysis collects the same analysis results of the dictionary correction impact analysis for each processing target sentence, selects one representative processing target sentence for each group, The process of creating an abstract of the impact analysis is performed by arranging the impact analysis result and the language processing result of the representative processing target sentence together with the number of processing target sentences belonging to the group.

  Further, the dictionary correction management means 21 performs a process of transmitting the impact analysis abstract created by the impact analysis abstract creation means 28 to the output device 4.

  Next, operation | movement of the dictionary reinforcement | strengthening assistance system 140 in the 12th Embodiment of this invention is demonstrated with reference to FIG.

  In this example, in the dictionary strengthening support process, the dictionary strengthening support system 140 performs the above-described steps S101 to S106.

  In the present embodiment, in step S106, the dictionary modification impact analysis unit 25 stores the language processing result based on the master dictionary and the language processing result based on the temporary registration dictionary in the impact analysis table storage unit 33. According to the impact analysis table, and output the analysis results.

  Further, the dictionary correction influence analysis means 25 calculates a dictionary correction influence score for each analysis result, and transmits it to the impact analysis abstract creation means 28.

  Upon receiving the analysis result of each processing target sentence from the dictionary correction influence analysis means 25, the impact analysis abstract creation means 28 groups each of the processing target sentences having the same dictionary analysis influence analysis result, and groups each of them. Determine the number of processing target sentences that belong to a group, determine the processing target sentence of each group, and arrange the impact analysis results and language processing results of the processing target sentences of each group together with the number of processing target sentences that belong to that group. An analysis abstract is created (step S1101), and the created impact analysis abstract is transmitted to the dictionary correction management means 21.

  The dictionary correction management unit 21 transmits the abstract of the impact analysis result received from the abstract creation unit 28 of the impact analysis to the output device 4, and causes the output device 4 to present the abstract of the impact analysis result (step S1102).

  As described above, in the twelfth embodiment, processing target sentences giving the same analysis result are grouped together, and the language processing result and the influence analysis result of the representative processing target sentence are presented as an abstract. Since the configuration is adopted, it is possible to reduce the amount of information to be presented while covering patterns of influence of dictionary correction. For this reason, even a dictionary worker who is not familiar with the language processing system and is not skilled in dictionary work can easily analyze the influence of dictionary correction on a large amount of analysis result differences and determine dictionary correction registration. Can do.

Embodiment 13 FIG.
Next, a thirteenth embodiment of the present invention will be described.
Since the dictionary strengthening support system in the thirteenth embodiment has the same configuration as the dictionary strengthening support system 140 shown in FIG. 14 described above, the dictionary strengthening support system 140 will be described with reference to the dictionary strengthening support system 140 in FIG. Detailed description thereof is omitted.

  In the present embodiment, the abstract creation means 28 of the impact analysis causes a processing target sentence and side effects that are likely to be improved by the dictionary correction in the impact analysis result of dictionary correction for each processing target sentence. Select processing target statements that have obtained analysis results that are particularly noteworthy in judging the registration of dictionary corrections from the effects of dictionary corrections, such as processing target statements that are likely to exist, processing target statements that do not cause differences, and this Performs the process of creating an abstract of the impact analysis by arranging the impact analysis results of the target sentence and the language processing results.

  Further, the dictionary correction management means 21 performs a process of transmitting the impact analysis abstract created by the impact analysis abstract creation means 28 to the output device 4.

  Next, operation | movement of the dictionary reinforcement | strengthening assistance system 140 in the 13th Embodiment of this invention is demonstrated with reference to FIG.

  In this example, in the dictionary strengthening support process, the dictionary strengthening support system 140 performs the above-described steps S101 to S106.

  In the present embodiment, in step S106, the dictionary modification impact analysis unit 25 stores the language processing result based on the master dictionary and the language processing result based on the temporary registration dictionary in the impact analysis table storage unit 33. According to the impact analysis table, and output the analysis results.

  Further, the dictionary correction influence analysis means 25 calculates a dictionary correction influence score for each analysis result and outputs it to the impact analysis abstract creation means 28.

  Upon receiving the analysis result of each processing target sentence from the dictionary correction influence analysis unit 25, the impact analysis abstract creation unit 28 improves the analysis result by the dictionary correction in the influence analysis result of the dictionary correction for each processing target sentence. Analyzes that require particular attention in determining the registration of dictionary corrections based on the effects of dictionary corrections, such as processing target sentences that are likely to have side effects, processing target sentences that are likely to have side effects, and processing target sentences that do not cause differences The processing target sentence from which the result is obtained is selected, the effect analysis result of the processing target sentence and the language processing result are arranged to create an abstract of the impact analysis (step S1101), and this is output to the dictionary correction management means.

  The dictionary correction management means 21 transmits the abstract of the impact analysis result to the output device 4, and causes the output device 4 to present the abstract of the impact analysis result (step S1102).

  As described above, in the thirteenth embodiment, a processing target sentence that has obtained an analysis result that should be particularly noted in determining registration of dictionary correction from the influence of dictionary correction is selected, and this processing target sentence is selected. The analysis results and the language processing results are arranged side by side as an abstract of the impact analysis, so the sentence to be processed, its analysis results, and the language processing results are presented, focusing on information useful for judging the registration of dictionary corrections. can do. For this reason, even a dictionary worker who is not familiar with the language processing system and is not skilled in dictionary work can easily analyze the influence of dictionary correction on a large amount of analysis result differences and determine dictionary correction registration. Can do.

Embodiment 14 FIG.
Next, a fourteenth embodiment of the present invention will be described.
FIG. 16 is a block diagram showing an example of the configuration of the dictionary strengthening support system 150 according to the fourteenth embodiment of the present invention. In FIG. 16, parts having the same configuration as that of the dictionary strengthening support system 100 shown in FIG. 1 described above are assigned the same reference numerals and detailed description thereof is omitted.

  As shown in FIG. 16, the dictionary strengthening support system 150 includes an input device 1, a data processing device 2 </ b> D, a storage device 3, and an output device 4.

  The data processing device 2D includes a dictionary correction management unit 21, a dictionary editing unit 22, a language processing unit 23, a processing target sentence selection unit 24, a dictionary correction influence analysis unit 25, and a dictionary registration determination unit 29.

  That is, the dictionary strengthening support system 150 is configured such that the data processing device 2 further includes a dictionary registration determination unit 29 as compared with the dictionary strengthening support system 100 described above.

  The dictionary registration determining unit 29 has a function of determining whether or not to register a dictionary correction in the master dictionary based on the influence analysis result of the processing target.

  Next, the operation of the dictionary strengthening support system 150 in the fourteenth embodiment of the present invention will be described. FIG. 17 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 150.

  As shown in FIG. 17, in this example, in the dictionary strengthening support process, the dictionary strengthening support system 150 performs the processes of steps S101 to S106 described above.

  In the present embodiment, in step S106, the dictionary modification impact analysis unit 25 stores the language processing result based on the master dictionary and the language processing result based on the temporary registration dictionary in the impact analysis table storage unit 33. According to the impact analysis table, and output the analysis results.

  Further, the dictionary correction influence analysis means 25 outputs the influence analysis result of each processing target sentence to the dictionary registration determination means 29.

  The dictionary registration determining unit 29 stores the dictionary correction input in the dictionary correction input (step S <b> 101) in the master dictionary storage unit 31 based on the influence analysis result of each processing target sentence from the dictionary correction influence analyzing unit 25. It is determined whether or not it should be reflected in the master dictionary (step S1401), and the dictionary registration determination result is transmitted to the dictionary correction management means 21.

  The dictionary correction management unit 21 sends the dictionary correction registration determination result from the dictionary registration determination unit 29 to the output device 4 together with the analysis result from the dictionary correction influence analysis unit 25 and the language processing result of the processing target sentence from the language processing unit 23. Send.

  The output device 4 performs a process of presenting the language processing result, the analysis result, and the dictionary registration determination result (step S1402).

  As described above, in the fourteenth embodiment, based on the influence analysis result of each processing target sentence, it is determined whether or not to reflect in the master dictionary for dictionary correction, and the determination result is presented. Therefore, the dictionary operator who has presented the determination result can be used as a reference for determining whether the dictionary correction should be reflected in the master dictionary, or can be used as a reference when analyzing the influence analysis result. Therefore, even a dictionary worker who is not familiar with the language processing system and is not skilled in dictionary work can easily analyze the influence of dictionary correction from a large amount of language processing results and determine whether to make dictionary correction registration. be able to.

Embodiment 15 FIG.
Next, a fifteenth embodiment of the present invention is described.
Since the dictionary strengthening support system in the fifteenth embodiment has the same configuration as the dictionary strengthening support system 150 shown in FIG. 16, it will be described with reference to the dictionary strengthening support system 150 in FIG. Detailed description thereof is omitted.

  In the present embodiment, the input device 1 performs a process of transmitting the dictionary registration determination result input by the operator's operation to the dictionary correction management means 21.

  If the dictionary registration determination result input by the input device 1 is to be registered, the dictionary editing unit 22 stores the dictionary editing result in the master dictionary stored in the master dictionary storage unit 31. Process to reflect. For example, the temporary registration dictionary stored in the temporary registration dictionary storage unit 32 is stored in the master dictionary storage unit 31 as a new master dictionary.

  Next, the operation of the dictionary strengthening support system 150 in the fifteenth embodiment of the present invention will be described. FIG. 18 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 150.

  As shown in FIG. 18, in this example, in the dictionary strengthening support process, the dictionary strengthening support system 150 performs the processes of steps S101 to S106, steps S1401, and S1402 described above, as in the fourteenth embodiment. Do.

  Next, the input device 1 transmits the dictionary registration determination result input by the dictionary operator's operation to the dictionary correction management means 21.

  When the dictionary registration determination result is received, the dictionary correction management means 21, when the dictionary registration determination result indicates that it is registered (Y in step S1501), the dictionary editing means reflects the request for reflecting the dictionary correction contents in the master dictionary. 22 to send.

  When receiving the reflection request, the dictionary editing unit 22 reflects the dictionary correction information stored in the temporary registration dictionary storage unit 32 in the master dictionary of the master dictionary storage unit 31 (step S1502).

  As described above, in the fifteenth embodiment, the correction can be immediately reflected in the master dictionary according to the dictionary correction registration judgment by the dictionary operator.

Embodiment 16 FIG.
Next, a sixteenth embodiment of the present invention will be described.
The dictionary strengthening support system in the sixteenth embodiment has the same configuration as the dictionary strengthening support system 150 shown in FIG. 16, and will be described with reference to the dictionary strengthening support system 150 in FIG. Detailed description thereof is omitted.

  In the present embodiment, the dictionary correction management unit 21 stores the dictionary editing result in the master dictionary storage unit 31 when it is determined that the correction should be reflected according to the dictionary correction registration determination result sent from the dictionary registration determination unit 29. The process which reflects in the master dictionary memorize | stored in is performed.

  Next, the operation of the dictionary strengthening support system 150 in the sixteenth embodiment of the present invention will be described. FIG. 19 is a flowchart illustrating an example of dictionary strengthening support processing executed by the dictionary strengthening support system 150.

  As shown in FIG. 19, the dictionary strengthening support process in the present embodiment is a process excluding the process in step S1402 in the fifteenth embodiment described above.

  In this embodiment, the dictionary registration determination unit 29 transmits the result of the dictionary registration determination in step S1401 to the dictionary correction management unit 21.

  The dictionary correction management unit 21 performs dictionary correction registration determination according to the dictionary registration determination result from the dictionary registration determination unit 29 (step S1501). That is, if the dictionary registration determination result from the dictionary registration determination unit 29 is a determination to register a dictionary, the dictionary correction management unit 21 transmits a dictionary correction registration request to the dictionary editing unit 22.

  When receiving the reflection request, the dictionary editing unit 22 reflects the dictionary correction information stored in the temporary registration dictionary storage unit 32 in the master dictionary of the master dictionary storage unit 31 (step S1502).

  As described above, in the sixteenth embodiment, since the dictionary correction is automatically registered in the master dictionary according to the dictionary correction registration determination, in the processing target sentence obtained by the dictionary correction If the influence on the language processing result is obviously not a problem, correction can be registered without requiring the dictionary operator's confirmation, so that the burden on the dictionary operator can be reduced.

  Although not specifically mentioned in the above-described embodiment, each process executed in the system 100, 110, 120, 130, 140, 150 is a control program (dictionary strengthening support) installed in the system 100. Program). This control program is a dictionary strengthening support program for assisting correction work for strengthening a word dictionary in which word information including words, which is a master dictionary used for natural language processing, is registered. From the step of creating a temporary registration dictionary obtained by editing the word dictionary based on the dictionary editing information indicating the correction instruction content of the word dictionary, and the processing target candidate sentence that is a candidate for the processing target of the natural language processing, A step including selecting a sentence including a word to be corrected in the dictionary editing information as a processing target sentence, and executing natural language processing based on each of the master dictionary and the temporary registration dictionary on the selected processing target sentence The processing result of the natural language processing on the processing target sentence based on the step, the master dictionary, and the processing target sentence based on the temporary registration dictionary. Comparing the processing results of natural language processing and analyzing the influence of dictionary editing according to the correction instruction contents indicated by the dictionary editing information on a predetermined viewpoint, and analyzing the results together with the processing results of natural language processing A program for executing the presented step.

Next, a first specific example of the present invention will be described.
Here, with reference to FIG. 1 and FIG. 2, a specific example in the above-described first embodiment will be described.

  In this example, it is assumed that the language processing unit 23 performs morpheme analysis processing that decomposes the processing target sentence into morpheme units each including a word or a particle as a language process. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described.

  When it is input that the dictionary operator newly registers “thought” (adverb) by the dictionary operator's input operation, the input device 1 newly registers “thought” (adverb) to the dictionary correction management means 21. The dictionary correction information indicating is transmitted.

  When the dictionary correction management unit 21 receives the dictionary correction information from the input device 1 (step S101), the dictionary editing unit 22 generates a temporary registration dictionary in which the correction contents indicated by the dictionary correction information are reflected in the master dictionary. Send a dictionary editing request to instruct.

  When the dictionary editing request is received, the dictionary editing means 22 generates a temporary registration dictionary in which “thought” (adverb) is added to the master dictionary and stores it in the temporary registration dictionary storage unit 32 (step S102).

  Next, the dictionary operator operates the input device 1 and, for example, as shown in FIG. 20, as a sentence including the surface layer of the added word “thought” (adverb), the input sentence 1 “each thought…” It is assumed that the input sentence 2 “A work of your choice” is entered.

  When the input sentence 1 and the input sentence 2 from the input device 1 are received (step S103), the dictionary correction management unit 21 instructs the processing target sentence selecting unit 24 to use the input sentence 1 and the input sentence 2 as processing target candidate sentences. At the same time, a request for selecting a processing target sentence is transmitted.

  When receiving the processing target sentence selection request from the dictionary correction management unit 21, the processing target sentence selecting unit 24 includes both the input sentence 1 and the input sentence 2 including the surface layer of the word “thought” to be corrected. Both 1 and the input sentence 2 are selected as processing target sentences (step S104), and the selection result is transmitted to the dictionary correction management means 21.

  Upon receiving the processing target sentence selection result, the dictionary correction management unit 21 issues a language processing request to the language processing unit 23 for the input sentence 1 and the input sentence 2 that are the processing target sentences.

  Upon receiving the language processing request, the language processing means 23 performs language processing on the input sentence 1 and the input sentence 2 based on the master dictionary and the temporary registration dictionary, respectively (step S105).

  In this example, a morphological analysis result as shown in FIG. 20 is obtained as the language processing result. That is, for the input sentence 1, the same result is obtained in the master dictionary and the temporary registration dictionary, and for the input sentence 2, different results are obtained in the master dictionary and the temporary registration dictionary. As shown in FIG. 20, in the morphological analysis for the input sentence 1, rather than the connection of “thought” (adverb) + “ni” (particle), “thought” (noun) + “thought” (noun) + It is assumed that the connection of “Ni” (particle) is processed with a higher priority.

  The dictionary correction management unit 21 transmits the result of the language processing by the language processing unit 23 and the analysis request to the dictionary correction influence analysis unit 25.

  When the analysis request is received, the dictionary modification influence analysis means 25 compares the language processing result based on the master dictionary with the language processing result based on the temporary registration dictionary, and compares and analyzes the difference for each processing target sentence (step). S106).

  In this example, the dictionary correction influence analysis means 25 uses the items shown in each item of “phenomenon seen in the difference” shown in FIG. 21 as a viewpoint for judging the influence of the dictionary correction, and processes the processing target for each phenomenon in each viewpoint. Compare and analyze differences between sentences.

  The viewpoints exemplified as “phenomenon seen in the difference” include “disappearance of unregistered words”, “decrease of morpheme number”, “generation of corrected (registered) words”, “ “No difference”, “Generation of morphemes that are not internal (external) inclusive”, “Generation of functional words”, “Disappearance of corrected (registered) words”, and “Generation of unregistered words” are exemplified.

  Here, the influence of the dictionary correction in each processing target sentence unit is analyzed based on a subjective evaluation value called a dictionary correction influence score. In this example, as an example of the dictionary correction influence score, as shown in FIG. 21, the magnitude of the influence is expressed as an absolute value, and among the phenomena seen in the difference, a phenomenon where the influence is an improvement is positive or negative. Use a real value that represents the phenomenon negatively.

  It should be noted that the dictionary correction influence score of each processing target sentence is obtained by summing up the dictionary correction influence scores for each phenomenon found in the difference (each phenomenon corresponding to each item of “Phenomenon found in difference” in FIG. 21). It is done.

  As shown in FIG. 20, the input sentence 1 has no change in the language processing result before and after the dictionary correction, so “no difference” corresponds to the phenomenon seen in the difference set in FIG. Therefore, the dictionary correction influence score of the input sentence 1 is “+0.0”.

  For the input sentence 2, as shown in FIG. 20, the morpheme of the word “thought” to be registered is newly generated by correcting the dictionary, and therefore, among the phenomena seen in the difference set in FIG. “Generating corrected (registered) words” corresponds. Further, for the input sentence 2, since the morpheme number is decreased by one due to the dictionary correction, “decrease of the morpheme number” among the phenomena seen in the difference set in FIG. 21 also corresponds. Therefore, the dictionary correction influence score of the input sentence 1 is (+0.3) + (+ 0.6) = “+ 0.0”.

  The dictionary correction influence analysis unit 25 transmits the derived dictionary correction influence score to the dictionary correction management unit 21 together with the language processing results of the input sentence 1 and the input sentence 2.

  The dictionary correction management means 21 transmits the language processing results of the input sentence 1 and the input sentence 2 to the output device 4 together with the dictionary correction influence score.

  For example, as shown in FIG. 22, the output device 4 presents the language processing results of the input sentence 1 and the input sentence 2 and the dictionary correction influence score by displaying them on the screen (step S107).

  As shown in FIG. 22, the output device 4, in this embodiment, each input sentence 1, 2 as a processing target sentence, each language processing result before and after correction in each processing target sentence, dictionary correction influence score, The phenomenon seen in the difference between the language processing results before and after the correction in each processing target sentence is displayed on the screen by being classified for each processing target sentence.

Next, a second specific example of the present invention will be described.
Here, with reference to FIG. 3 and FIG. 4, the specific Example in 2nd Embodiment mentioned above is demonstrated.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described.

  Furthermore, in this example, it is assumed that example sentences 1 to n shown in FIG. 23 are stored in advance in the example sentence storage unit 34.

  In the dictionary strengthening support process, the dictionary strengthening support system 110 first executes the processes of steps S101 to S102 as in the first embodiment.

  Next, upon receiving a processing target sentence selection request from the dictionary correction management unit 21, the processing target sentence selecting unit 24 selects the correction target word “from the example sentences 1 to n stored in the example sentence storage unit 34. An example sentence including the surface layer of “thought” (adverb) is selected as a processing target sentence (step S201).

  In this example, it is assumed that, in step S201, the example sentence 1 “presenting the work of thought” and the example sentence 2 “each device is devised” are selected as the processing target sentences. .

  The processing target sentence selection unit 24 transmits the example sentence 1 and the example sentence 2 selected as the processing target sentence to the dictionary correction registration management unit 21 as a selection result.

  Upon receiving the selection result of the processing target sentence, the dictionary correction management unit 21 transmits a language processing request for the example sentence 1 and the example sentence 2 as the processing target sentence to the language processing unit 23.

  And the dictionary reinforcement | strengthening assistance system 110 performs the process of step S105-step S107 similarly to Example 1 mentioned above.

Next, a third specific example of the present invention will be described.
Here, with reference to FIG. 5 and FIG. 6, a specific example in the third embodiment described above will be described.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described.

  Furthermore, in this example, it is assumed that example sentences 1 to n shown in FIG. 23 are stored in advance in the example sentence storage unit 34.

  In the dictionary strengthening support process, the dictionary strengthening support system 110 first executes the processes of steps S101 to S102 as in the first embodiment.

  Next, it is assumed that the dictionary operator operates the input device 1 and inputs the input sentence “how to feel each person feels good” as a sentence including the surface layer of the word “thought” (adverb) to be added. .

  When the input sentence from the input device 1 is received (step S103), the dictionary correction management means 21 transmits an example sentence registration request to the example sentence registration means 26.

  Upon receiving the example sentence registration request, as shown in FIG. 24, the example sentence registration means 26 performs a process of additionally registering the input sentence “how to feel each person feels good” as the example sentence n + 1 in the example sentence storage unit 34 (step S301). ).

  That is, the registration content of the example sentence storage unit 34 is updated from the content shown in FIG. 23 to the content shown in FIG. 24 by the registration process in step S301.

  Next, the dictionary correction management unit 21 transmits a processing target sentence selection request to the processing target sentence selection unit 24.

  When receiving the processing target sentence selection request from the dictionary correction management means 21, the processing target sentence selection means 24 selects the word “thought” (adverb) to be corrected from the example sentences stored in the example sentence storage unit 34. An example sentence including the surface layer (example sentence 1, example sentence 2, example sentence n + 1 shown in FIG. 24) is selected as a processing target sentence (step S201), and the selection result is transmitted to the dictionary correction management means 21.

  Upon receiving the selection result of the processing target sentence, the dictionary correction management unit 21 issues a language processing request to the language processing unit 23 for the example sentence 1, the example sentence 2 and the example sentence n + 1 that are the processing target sentences.

  And the dictionary reinforcement | strengthening assistance system 110 performs the process of step S105-step S107 similarly to Example 1 mentioned above.

Next, a fourth specific example of the present invention will be described.
Here, with reference to FIGS. 3 and 7, a specific example in the above-described fourth embodiment will be described.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described. In this example, it is assumed that no example sentence is registered in advance in the example sentence storage unit 34.

  In the dictionary strengthening support process, the dictionary strengthening support system 110 first executes the processes of steps S101 to S102 and step S201 as in the second embodiment.

  In this example, the input sentence is not transmitted from the dictionary correction management means 21, and the example sentence is not registered in the example sentence storage unit 34, and the processing target sentence is not selected in step S201 (N in step S401). The processing target sentence selection unit 24 notifies the dictionary correction management unit 21 that there is no processing target sentence among the selection candidate sentences.

  Upon receiving notification that the processing target sentence does not exist, the dictionary correction management means 21 requests the output device 4 to input the processing target sentence in order to instruct the dictionary operator to input the processing target sentence. An input request instruction for presenting is made (step S402).

  In response to the input request instruction, the output device 4 displays, for example, an input request sentence “Please input a sentence including the newly registered word“ thought ”” in order to prompt the input of a sentence to be subjected to language processing. Output to the screen.

  In response to the input request, when the sentence “how to feel is perceived by each person” is input from the input device 1 (Y in step S403), the dictionary correction management means 21 causes the input sentence “how to perceive each person is perceived. And a processing target sentence selection request are transmitted to the processing target sentence selecting means 24. When receiving the request for selecting the processing target sentence, the processing target sentence selecting unit 24 selects the processing target sentence from the input sentences (step S104). Then, the process proceeds to step S401.

  Thereafter, since the processing target sentence is selected in step S104 (Y in step S401), the processes in steps S105 to S107 are executed.

Next, a fifth specific example of the present invention will be described.
Here, with reference to FIG. 5 and FIG. 8, a specific example in the above-described fifth embodiment will be described.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described. In this example, it is assumed that no example sentence is registered in advance in the example sentence storage unit 34.

  In the dictionary strengthening support process, the dictionary strengthening support system 120 first executes the processes of steps S101 to S103, as in the third embodiment.

  That is, in step S103, as in the third embodiment described above, it is assumed that the input sentence “how to feel is acceptable for each person” is received from the input device 1.

  When the input sentence from the input device 1 is received (step S103), the dictionary correction management unit 21 transmits a processing target sentence selection request to the processing target sentence selecting unit 24.

  Next, when the processing target sentence selection unit 24 receives the processing target sentence selection request from the dictionary correction management unit 21, no example sentence is registered in the example sentence storage unit 34. Only “Yes” is selected as a processing target sentence (step S104, step S201).

  When the processing target sentence is selected, the processing target sentence selecting unit 24 adds origin information indicating that the input origin is added to the selected processing target sentence “how to feel for each person”, and adds the origin information. The target sentence is transmitted to the dictionary correction management means 21.

  The dictionary correction management unit 21 confirms the origin information added to the processing target sentence received from the processing target sentence selection unit 24, and the received processing target sentence does not have an example sentence origin (N in Step S501). An example sentence registration request is sent to the example sentence registering means 26 to register the input sentence “how to feel each person may think” as an example sentence (step S502).

  When the example sentence registration request is received, the example sentence registration unit 26 performs a process of registering the input sentence “how to feel each person feels good” accepted in step S103 as an example sentence in the example sentence storage unit 34 (step S301). Thereafter, the processes of steps S105 to S107 described above are performed.

Next, a sixth specific example of the present invention will be described.
Here, with reference to FIG. 5 and FIG. 9, a specific example in the above-described sixth embodiment will be described.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described.

  Furthermore, in this example, it is assumed that example sentences 1 to n shown in FIG. 23 are stored in advance in the example sentence storage unit 34.

  In the dictionary strengthening support process, the dictionary strengthening support system 120 first executes the processes of steps S101 to S103, as in the third embodiment.

  That is, in step S103, as in the third embodiment described above, it is assumed that the input sentence “how to feel is acceptable for each person” is received from the input device 1.

  When the input sentence from the input device 1 is received (step S103), the dictionary correction management unit 21 transmits a processing target sentence selection request to the processing target sentence selecting unit 24.

  Next, when the processing target sentence selection unit 24 receives the processing target sentence selection request from the dictionary correction management unit 21, the processing target sentence selection unit 24 and the example sentence 1 to example n registered in the example sentence storage unit 34 and the input sentence “how to feel is human. The sentence (in this example, example sentence 1, example sentence 2, input sentence) including the surface layer of the word “idea” (adverb) to be added is selected as a process target sentence from “Everything you want” (step S104, step S201).

  When the processing target sentence is selected, the processing target sentence selecting unit 24 adds origin information indicating that the input sentence originates to the input sentence “how to feel each person may think” as the processing target sentence. Origin information indicating that it is the origin of the example sentence is added to each of the example sentence 1 and the example sentence 2 selected as the target sentence, and each processing target sentence to which the origin information is added is transmitted to the dictionary correction management means 21.

  Next, upon receiving a language processing request for inputting the example sentence 1, example sentence 2 and the input sentence from the dictionary correction management means 21, the language processing means 23 applies the master dictionary and the temporary sentence to the example sentence 1, example sentence 2 and the input sentence. Each language process is executed based on the registered dictionary (step S105), and a morphological analysis result as shown in FIG. 25 is obtained.

  When the language processing is finished, the language processing means 23 sends the processing result to the dictionary correction influence analyzing means 25.

  The dictionary correction influence analysis means 25 compares the language processing result based on the master dictionary with the language processing result based on the temporary registration dictionary, and the difference is found in “phenomenon seen in the difference” in FIG. 21 for each processing target sentence. A comparative analysis is performed for each phenomenon (step S106).

  Here, it is assumed that the influence of the dictionary correction for each processing target sentence is analyzed according to a subjective evaluation value called a dictionary correction influence score.

  As an example of the dictionary correction influence score, as shown in FIG. 21, a real value is used in which the magnitude of the influence is expressed as an absolute value, the influence is an improvement, and the influence is an improvement. The dictionary correction influence score of each target sentence is obtained by summing up each dictionary correction influence score set in advance corresponding to each phenomenon found in the difference ("Phenomenon seen in the difference" in FIG. 21).

  In this example, in step S106, the dictionary correction influence analysis means 25 performs a process of deriving a dictionary correction influence score for each processing target sentence using the influence analysis table shown in FIG.

  Specifically, as shown in FIGS. 20 and 21, for the example sentence 1 “each thought”, “no difference” corresponds to the phenomenon seen in the difference, so the dictionary correction influence score is +0. 0. In addition, for example sentence 2 “The work of thought ...” and the input sentence “... may be thoughtful”, the phenomenon of “decrease in morpheme” and “corrected (registered) words” are the phenomena seen in the difference, respectively. Since “Generate” corresponds, the dictionary correction influence score is +0.9.

  The dictionary correction influence analysis means 25 derives the dictionary correction influence score for each processing target sentence as described above, and the dictionary correction influence score thus derived, together with the language processing results of the example sentence 1 and the example sentence 2, the dictionary correction management means 21. Send to.

  Thereafter, the dictionary correction management means 21 obtains the processing sentence from the input sentence based on the processing sentence and origin information from the processing sentence selection means 24 and the analysis result from the dictionary correction influence analysis means 25. It is confirmed whether or not the analyzed result matches any of the analysis results for the sentence to be processed originating in the example sentence (step S601).

  If it matches any of the analysis results for the sentence to be processed originating in the example sentence (N in step S601), the dictionary strengthening support system 120 performs the process in step S107 described above.

  Here, as a criterion for determining whether or not the analysis results are different, for example, among the “phenomenon seen in the difference” shown in FIG. Can be used for reference.

  In this case, both the example sentence 2 and the input sentence correspond to “decrease in the number of morphemes” and “generation of corrected (registered) words”, so that there is an example sentence that matches the analysis result of the input sentence. In such a case, the example sentence registration request for the input sentence is not transmitted.

  Further, as a criterion for determining whether or not the analysis results are different, for example, it can be based on whether or not the surface layer of the correction target word matches the morpheme string including one morpheme before and after that.

  In this case, in the example sentence 2 and the input sentence, the analysis results in the master dictionary “/ mind (noun) / mind (noun) / no (particle) /” and “/ mind (noun) / mind (noun) / in (particle) )) ”, Analysis result in temporary registration dictionary“ / thought (adverb) / no (help) / ”and“ / thought (adverb) / de (help) / ”did not match, and entered Example sentence 1 Since sentences do not match, an example sentence registration request for an input sentence is transmitted to the dictionary registration means 26.

  It is determined whether or not the analysis result is different according to the discrimination criterion as described above, and if it does not match any of the analysis results for the sentence to be processed originating in the example sentence (Y in step S601), the dictionary correction management means 21 An example sentence registration request is sent to the registration means 26 to give an instruction to register an input sentence that is an analysis result that does not match any of the analysis results for the sentence to be processed originating in the example sentence (step S502).

  When the example sentence registration request is received, the example sentence registering means 26 outputs an input sentence that becomes an analysis result that does not match any of the analysis results for the sentence to be processed among the input sentences received in step S103, to the example sentence storage unit. A process of registering as an example sentence in 34 is performed (step S301). Then, the process of step S107 mentioned above is performed.

Next, a seventh specific example of the present invention will be described.
Here, with reference to FIG. 1 and FIG. 10, a specific example in the above-described seventh embodiment will be described.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “mail software” (noun), which is a dictionary correction target word, to the dictionary will be described.

  In the dictionary strengthening support process, the dictionary strengthening support system 120 first executes the processes of steps S101 to S102 as in the first embodiment. However, in this example, a dictionary registration of the word “mail software” (noun) is input from the input device 1, and a temporary registration dictionary in which “mail software” (noun) is added to the contents of the master dictionary is a temporary registration dictionary storage unit 32. Is remembered.

  Next, in this example, the dictionary operator operates the input device 1 to input a sentence including the surface layer of the word “mail software” (noun) to be added and a language processing result for the sentence.

  In this example, as shown in FIG. 26, as a sentence including the surface layer of the word “mail software” (noun) to be added, input sentence 1 “install mail software...” And input sentence 2 “mail software share /・ ・ 」Is entered. As the language processing result, as shown in FIG. 26, the language processing result “/ mail (change) / software (noun) / no (particle) / share (noun) /... The

  In response to the input operation by the dictionary operator, the input device 1 transmits the input sentence 1, the input sentence 2, and the language processing result of the input sentence 2 shown in FIG.

  The dictionary correction management unit 21 receives the input sentence 1, the input sentence 2, and the language processing result of the input sentence 2 from the input device 1 (step S701), and receives the received input sentence to the processing target sentence selecting unit 24. 1. The input sentence 2 and the language processing result of the input sentence 2 are transmitted, and a processing target sentence selection request is transmitted.

  When receiving the processing target sentence selection request, the processing target sentence selecting unit 24 selects the input sentence 1 and the input sentence 2 including the surface layer of the dictionary correction target word as the processing target sentence, and the language of the received input sentence 2 The processing result is selected as the reference language processing result (step S104).

  When the processing target sentence and the reference language processing result are selected, the processing target sentence selecting unit 24 selects the input sentence 1 and the input sentence 2 selected as the processing target sentence and the language processing result of the input sentence 2 selected as the reference language processing result. To the dictionary correction management means 21.

  When the input sentence 1 and the input sentence 2 are received, the dictionary correction management means 21 requests the language processing means 23 to create the language processing results of the input sentence 1 and the input sentence 2 based on the master dictionary and the temporary registration dictionary, respectively. Send.

  When the language processing result creation request is received, the language processing unit 23 performs language processing on the input sentence 1 and the input sentence 2 (step S105). In this example, a morphological analysis result as shown in FIG. 27 is obtained in step S105.

  Next, the language processing unit 23 transmits the language processing result shown in FIG. 27 to the dictionary correction management unit 21.

  Upon receiving the language processing result, the dictionary correction management means 21 dictionary the received language processing result (see FIG. 27) and the reference language processing result of the input sentence 2 (language processing result of the input sentence 2 shown in FIG. 26). It transmits to the correction influence analysis means 25.

  When the language processing result and the reference language processing result of the input sentence 2 are received, the dictionary modification influence analysis means 25 performs the language processing result based on the master dictionary, the language processing result based on the temporary registration dictionary, and the reference language of the input sentence 2 The processing result is compared, and the difference for each processing target sentence is compared and analyzed for each phenomenon as shown in “Phenomenon found in difference” in FIG. 21 (step S106).

  Here, it is assumed that the influence of the dictionary correction for each processing target sentence is analyzed according to a subjective evaluation value called a dictionary correction influence score.

  As an example of the dictionary correction influence score, as shown in FIG. 21, a real value is used in which the magnitude of the influence is expressed as an absolute value, the influence is an improvement, and the influence is an improvement. The dictionary correction influence score of each target sentence is obtained by summing up each dictionary correction influence score set in advance corresponding to each phenomenon found in the difference ("Phenomenon seen in the difference" in FIG. 21).

  In this example, in step S106, the dictionary correction influence analysis means 25 performs a process of deriving a dictionary correction influence score for each processing target sentence using the influence analysis table shown in FIG.

  Specifically, as shown in FIGS. 21 and 27, for the input sentence 1 “install mail software...”, “Decrease in morpheme” and “corrected (registered) words as phenomena seen in the difference. Therefore, the dictionary correction influence score is +0.9. In addition, for the input sentence 2 “share of mail software ...”, “generation of corrected (registered) words” and “generation of morphemes that are not internal (external) inclusive relations” correspond to the phenomenon seen in the difference. The dictionary correction influence score is +0.0.

  Further, after the analysis is completed, the dictionary correction influence analysis means 25 obtains the dictionary correction score +0.9 of the input sentence 1, the dictionary correction score +0.0 of the input sentence 2, and the reference language processing result of the input sentence 2 as the master dictionary. To the dictionary correction management means 21 together with the respective language processing results of the input sentence 1 and the input sentence 2, that the language processing result of the temporary registration dictionary is not matched.

  The dictionary correction management means 21 is the result of processing each language of the input sentence 1 and the input sentence 2, the analysis result of the dictionary correction effect (the dictionary correction score of the input sentence 1 +0.9, the dictionary correction score of the input sentence 2 +0.0), And the comparison result between the language processing result of the input sentence 2 and the reference language processing result (the reference language processing result of the input sentence 2 matches the language processing result of the master dictionary and does not match the language processing result of the temporary registration dictionary ), For example, an output format as shown in FIG. 28 is created and transmitted to the output device 4.

  The output device 4 displays the screen based on the output format created by the dictionary correction management means 21, thereby processing each language of the input sentence 1 and the input sentence 2, analyzing the effect of dictionary correction, and the language of the input sentence 2. A comparison result between the processing result and the reference language processing result is presented (step S107).

  In this example, as shown in FIG. 28, based on the comparison result between the language processing result of the input sentence 2 and the reference language processing result, the “reference language processing” indicates that the corresponding dictionary correction has an adverse effect on the language processing. Regarding the result, “degrade” is presented.

Next, an eighth specific embodiment of the present invention will be described.
Here, with reference to FIGS. 3 and 4, a specific example in the above-described eighth embodiment will be described.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “mail software” (noun), which is a dictionary correction target word, to the dictionary will be described.

  Furthermore, in this example, it is assumed that example sentences 1 and 2 shown in FIG. 29 are stored in advance in the example sentence storage unit 34. Also, as shown in FIG. 29, it is assumed that the example sentence 2 is registered in the example sentence storage unit 34 with the reference language processing result added.

  In the dictionary strengthening support process, the dictionary strengthening support system 110 first executes the processes of steps S101 to S102 as in the seventh embodiment. In other words, a dictionary registration of the word “mail software” (noun) is input from the input device 1, and a temporary registration dictionary in which “mail software” (noun) is added to the contents of the master dictionary is stored in the temporary registration dictionary storage unit 32. I do.

  Next, upon receiving a processing target sentence selection request from the dictionary correction management means 21, the processing target sentence selection means 24 selects the correction target word “1” from the example sentences 1 to 2 stored in the example sentence storage unit 34. An example sentence including the surface layer of “mail software” (noun) is selected as a processing target sentence (step S201).

  In this example, the example sentence 1 and the example sentence 2 shown in FIG. 29 are selected as processing target sentences in step S201.

  The processing target sentence selection unit 24 transmits the example sentence 1 and the example sentence 2 selected as the processing target sentence to the dictionary correction registration management unit 21 as a selection result.

  Upon receiving the selection result of the processing target sentence, the dictionary correction management means 21 transmits to the language processing means 23 the example language 1, the example 2 and the reference language processing result added to the example 2 shown in FIG. Then, a language processing request is made for the example sentence 1 and the example sentence 2 that are the processing target sentences.

  And the dictionary reinforcement | strengthening assistance system 110 performs the process of step S105-step S107 similarly to Example 7 mentioned above.

Next, a ninth specific embodiment of the present invention will be described.
Here, a specific example of the ninth embodiment will be described with reference to FIGS. 11 and 12.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “mail software” (noun), which is a dictionary correction target word, to the dictionary will be described.

  Furthermore, in this example, it is assumed that example sentences 1 and 2 shown in FIG. 29 are stored in advance in the example sentence storage unit 34. Also, as shown in FIG. 29, it is assumed that the example sentence 2 is registered in the example sentence storage unit 34 with the reference language processing result added.

  In the dictionary strengthening support process, the dictionary strengthening support system 110 first executes the processes of steps S101 to S102 as in the seventh embodiment. In other words, a dictionary registration of the word “mail software” (noun) is input from the input device 1, and a temporary registration dictionary in which “mail software” (noun) is added to the contents of the master dictionary is stored in the temporary registration dictionary storage unit 32. I do.

  Next, in this example, the dictionary operator operates the input device 1 to input a sentence including the surface layer of the word “mail software” (noun) to be added and a language processing result for the sentence.

  In this example, as shown in FIG. 30, the same input sentence “install mail software...” As the example sentence 1 is input as a sentence including the surface layer of the word “mail software” (noun) to be added. Also, as the language processing result, as shown in FIG. 30, the language processing result “/ email (sa modification) / software (noun) / a (particle) / installation (sa modification) /... .

  In response to the input operation by the dictionary operator, the input device 1 transmits the input sentence and the language processing result of the input sentence to the data processing device 2.

  The dictionary correction management unit 21 receives the input sentence from the input device 1 and the language processing result of the input sentence (step S701), and sends the received input sentence and the input sentence language to the processing target sentence selecting unit 24. A processing result is transmitted and a processing target sentence selection request is made.

  Here, when the dictionary operator inputs an example sentence registration request for an input sentence from the input device 1, the dictionary correction management unit 21 sends an example sentence registration request for the input sentence (see FIG. 30) to the example sentence registration unit 26.

  The example sentence registering means 26 does not perform new registration and does not register the example sentence 1 in the example sentence storage unit 34 because the input sentence requested to register the example sentence is the same sentence as the example sentence 1 registered in advance (see FIG. 29). The storage position is transmitted to the dictionary correction management means 21.

  When the storage position of the example sentence 1 is received, the dictionary correction management unit 21 transmits the storage position of the example sentence 1 to the reference language processing result registration unit 27, and the language processing result of the input sentence (see FIG. 30) is transmitted to the example sentence storage unit. A registration request for registering in association with the example sentence 1 registered in 34 is transmitted.

  When the registration request is received, the reference language processing result registration unit 27 stores the language processing result of the input sentence in the example sentence storage unit 34 as the reference language processing result of the example sentence 1 as shown in FIG. 31 (step S901).

  Thereafter, similarly to the seventh embodiment described above, the processes of Steps S104 to S107 are executed.

Next, a tenth specific embodiment of the present invention will be described.
Here, with reference to FIG. 11 and FIG. 13, a specific example in the above-described tenth embodiment will be described.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “mail software” (noun), which is a dictionary correction target word, to the dictionary will be described.

  Furthermore, in this example, it is assumed that example sentences 1 and 2 shown in FIG. 29 are stored in advance in the example sentence storage unit 34. Also, as shown in FIG. 29, it is assumed that the example sentence 2 is registered in the example sentence storage unit 34 with the reference language processing result added.

  In the dictionary strengthening support process, the dictionary strengthening support system 130 first executes the processes of step S101 to step S103, step S104, S201, step S105 to step S107, as in the example described above.

  Then, the dictionary operator, for example, the analysis result “/ mail (change) / software (noun) / Is requested as a reference language processing result, the dictionary correction management means 21 sends an example sentence registration request for this sentence to the example sentence registration means 26.

  The example sentence registration unit 26 stores the example sentence 1 in the example sentence storage unit 34 without performing new registration because the sentence requested to register the example sentence is the same sentence as the example sentence 1 registered in advance (see FIG. 29). The position is transmitted to the dictionary correction management means 21.

  When the storage position of the example sentence 1 is received, the dictionary correction management means 21 transmits the storage position of the example sentence 1 to the reference language processing result registration means 27, and the analysis result “/ mail () requested to be registered by the dictionary operator. Registration for registering “example (1)” / software (noun) / (particle) / install (form) /... (See FIG. 28) in association with example sentence 1 registered in example sentence storage unit 34. Send a request.

  When the registration request is received, the reference language processing result registration unit 27 stores the language processing result of the input sentence in the example sentence storage unit 34 as the reference language processing result of the example sentence 1 as shown in FIG. 31 (step S1001).

Next, an eleventh specific embodiment of the present invention will be described.
Here, with reference to FIG. 14 and FIG. 15, a specific example in the eleventh embodiment will be described.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described.

  In the dictionary strengthening support process, the dictionary strengthening support system 140 first executes the processes of steps S101 to S106 as in the first embodiment.

  In this embodiment, in step S106, the dictionary modification impact analysis means 25 stores the language processing result based on the master dictionary and the language processing result based on the temporary registration dictionary in the impact analysis table storage unit 33. Based on the influence analysis table, the input sentence 1 and the input sentence 2 are analyzed, and the influence analysis result regarding the input sentence 1 and the input sentence 2 is output.

  Further, the dictionary correction influence analysis means 25 calculates a dictionary correction influence score for each analysis result, and outputs it to the influence analysis abstract creation means 28 for the input sentence 1 and the input sentence 2.

  Upon receiving the analysis results from the dictionary correction influence analysis means 25 (impact analysis results regarding the input sentence 1 and input sentence 2, dictionary correction influence score, language processing result (see FIG. 22)), the impact analysis abstract creation means 28 Based on the value of the dictionary correction influence score of the processing target sentence, processing for arranging the processing target sentences with the language processing result and the analysis result side by side is performed.

  In this example, the impact analysis abstract creating means 28 arranges the data in order of decreasing dictionary correction impact score, and creates, for example, an output format as shown in FIG. 32 as an abstract of the impact analysis result (step S1101). ). Then, the created abstract of the impact analysis result is transmitted to the dictionary correction management means 21.

  The dictionary correction management means 21 transmits the abstract of the impact analysis result to the output device 4, and causes the output device 4 to present the abstract of the impact analysis result as shown in FIG. 32 (step S1102).

Next, a twelfth specific embodiment of the present invention will be described.
Here, a specific example of the twelfth embodiment will be described with reference to FIGS. 14 and 15.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described.

  In the dictionary strengthening support process, the dictionary strengthening support system 140 first executes the processes of steps S101 to S106 as in the first embodiment.

  In this embodiment, in step S106, the dictionary modification impact analysis means 25 stores the language processing result based on the master dictionary and the language processing result based on the temporary registration dictionary in the impact analysis table storage unit 33. Based on the influence analysis table, the input sentence 1 and the input sentence 2 are analyzed, and the influence analysis result regarding the input sentence 1 and the input sentence 2 is output.

  Further, the dictionary correction influence analysis means 25 calculates a dictionary correction influence score for each analysis result and outputs it to the effect analysis abstract creation means 28 for the input sentence 1 and the input sentence 2.

  Upon receiving the analysis result of each processing target sentence from the dictionary correction influence analysis means 25 (impact analysis result on input sentence 1, input sentence 2, dictionary correction influence score, language processing result (see FIG. 22)), an impact analysis abstract is created. The means 28 is that the influence analysis result of the input sentence 1 is “reduction of morpheme number and generation of corrected (registered) word”, the influence analysis result of the input sentence 2 is “no difference”, and different analysis results are obtained. Therefore, these are divided into different groups, input sentence 1 and input sentence 2 are set as representative process target sentences of each group, and the impact analysis results and language processing results are arranged along with the number of cases, for example, An output format as shown in FIG. 33 is created as an abstract of the impact analysis result (step S1101).

  In other words, in step S1101, grouping the same impact analysis results into the same group, and information on the representative processing target sentences of each group are arranged side by side along with the number of processing target sentences belonging to the group. The process of creating the abstract is executed.

  Then, the impact analysis abstract creation means 28 transmits the created abstract of the impact analysis result to the dictionary correction management means 21.

  The dictionary correction management means 21 transmits the abstract of the impact analysis result to the output device 4, and causes the output device 4 to present the abstract of the impact analysis result as shown in FIG. 33 (step S1102).

Next, a thirteenth specific embodiment of the present invention will be described.
Here, a specific example of the thirteenth embodiment will be described with reference to FIGS. 14 and 15.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described.

  In the dictionary strengthening support process, the dictionary strengthening support system 140 first executes the processes of steps S101 to S106 as in the first embodiment.

  In this embodiment, in step S106, the dictionary modification impact analysis unit 25 stores the language processing result based on the master dictionary and the language processing result based on the temporary registration dictionary in the impact analysis table storage unit 33. Based on the influence analysis table, the input sentence 1 and the input sentence 2 shown in FIG. 20 are analyzed, and the influence analysis result regarding the input sentence 1 and the input sentence 2 is output.

  Further, the dictionary correction influence analysis means 25 calculates a dictionary correction influence score for each analysis result and outputs it to the effect analysis abstract creation means 28 for the input sentence 1 and the input sentence 2.

  Upon receiving the analysis result of each processing target sentence from the dictionary correction influence analysis means 25 (impact analysis result on input sentence 1, input sentence 2, dictionary correction influence score, language processing result (see FIG. 22)), an abstract of the influence analysis The creation means 28 is a processing target sentence that gives the maximum value of the dictionary correction influence score, a processing target sentence that is highly likely to have a side effect, as a processing target sentence that is likely to have an improved analysis result due to dictionary correction. Processing for which processing results that give the lowest dictionary correction impact score and processing target statements that do not produce a difference have obtained analysis results that are particularly noteworthy in judging registration of dictionary corrections from the effects of dictionary correction Select as target sentence.

  In the influence analysis result of the input sentence 1 and the input sentence 2 shown in FIG. 20, the input sentence 1 as the processing target sentence giving the maximum value of the dictionary correction influence score, the input sentence 2 as the processing target sentence giving the minimum value, and the difference The input sentence 2 is extracted as a processing target sentence that does not occur.

  The impact analysis abstract creating means 28 arranges the impact analysis result and the language processing result for each processing target sentence selected as the processing target sentence from which the notable analysis result is obtained, and outputs an output format as shown in FIG. Is created as an abstract of the impact analysis results (step S1101).

  Then, the impact analysis abstract creation means 28 transmits the created impact analysis result abstract to the dictionary correction management means 21.

  The dictionary correction management means 21 transmits the abstract of the impact analysis result to the output device 4, and causes the output device 4 to present the abstract of the impact analysis result as shown in FIG. 34 (step S1102).

Next, a fourteenth specific embodiment of the present invention will be described.
Here, with reference to FIGS. 16 and 17, a specific example of the above-described fourteenth embodiment will be described.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described.

  In the dictionary strengthening support process, the dictionary strengthening support system 150 first executes the processes of steps S101 to S106 as in the first embodiment.

  In this embodiment, in step S105, the language processing unit 23 creates a morpheme analysis result as shown in FIG. 20 for the input sentence 1 and the input sentence 2, and transmits the result to the dictionary correction influence analysis unit 25.

  In this embodiment, in step S106, the dictionary correction influence analysis means 25 compares the morphological analysis results as shown in FIG. 20, and the influence of the dictionary correction in each processing target sentence unit is called a dictionary correction influence score. Analyzes based on various evaluation values.

  As an example of the dictionary correction influence score, as shown in FIG. 35, a real value is used in which the magnitude of the influence is expressed as an absolute value, the influence is improved and the negative is expressed as positive.

  In this example, as shown in FIG. 35, the dictionary correction influence score is not set as a unique value as in Embodiments 1 to 13, but is set as a distribution function of the dictionary correction influence score.

  The merit of using the distribution function is that even if the value of the dictionary correction influence score that can be judged from the phenomenon seen in the difference has an error in estimating the influence, the magnitude of each error is handled by the variance (standard deviation). There is a point that can be.

Here, an example of a method for determining the value of the distribution function of the dictionary correction influence score will be described.
First, as shown in FIG. 35, a continuous distribution function fi (x) of a correction influence score x expected by dictionary correction is statistically estimated for each phenomenon seen in the difference, and a table of average μi and standard deviation σi is obtained. Create it in advance.

  Next, the distribution of the correction influence score for each phenomenon is synthesized, and the distribution function f (x) of the correction influence score x for the difference is estimated.

f (x) is estimated by a linear sum with weight wi from the distribution function fi (x) distribution function for each phenomenon as follows.
f (x) = ΣNi * fi (x)
(Where Ni is the normalized weight: wi / Σwj)

Then, the expected value μ and the standard deviation σ relating to the synthesized distribution function f (x) can be easily estimated as follows.
μ = ΣNi * μi
σ = [ΣNi * (σi ** 2 + μi ** 2) −μ ** 2] ** 1/2

  For example, for the input sentence 2 in FIG. 20, the phenomenon seen in the difference corresponds to “decrease in morpheme” and “generation of corrected (registered) words” in FIG. At this time, assuming that the distribution and weight of the correction influence score for each phenomenon are (μ1, σ1, N1) and (μ2, σ2, N2), the expected value and standard deviation of the synthesized distribution function are μ1 = 0.6, σ1 = 0.4, N1 = 0.5 / (0.5 + 0.3), and μ2 = 0.3, σ2 = 0.4, N2 = 0.3 / (0.5 + 0.3) ), The following is required.

μ = N1μ1 + N2μ2 = 0.49
σ = [N1 (σ1 ** 2 + μ1 ** 2) + N2 (σ2 ** 2 + μ2 ** 2) −μ ** 2] ** 1/2 = 0.43

  On the other hand, for the input sentence 1 in FIG. 20, since there is no difference between the analysis results of the master dictionary and the temporary registration dictionary, only “no difference” in FIG.

  From the above, the distribution of the correction influence score in the input sentence 1 is 0.0 ± 0.5, and the distribution of the correction influence score in the input sentence 2 is 0.49 ± 0.43.

  In step S106, these correction influence scores are transmitted to the dictionary registration determination means 29 together with the language processing results of the input sentence 1 and the input sentence 2.

  The dictionary registration determination unit 29 comprehensively evaluates the influence of dictionary correction on all the processing target sentences based on the comparison analysis result for each processing target sentence. The revision impact score distribution in each target sentence is synthesized, and the expected value and standard deviation are used as a guide. When the number of target sentences is N, the distribution of the correction influence score in the entire target sentence is estimated by the following equation.

  μ = Σμi / N, σ = [Σ (σi ** 2 + μi ** 2) / N−μ] ** 1/2

  For the two sentences of the input sentence 1 and the input sentence 2, μ = 0.25 and σ = 0.53.

  Here, since μ−σ <0 holds, it can be determined that there is a possibility that an improvement result as intended by the dictionary operator is not obtained or there is a side effect. Therefore, it is determined that dictionary registration should not be performed (step S1401).

  If μ−σ> 0 is satisfied, a message recommending registration is output. A threshold value may be further provided for the value of μ to classify the levels, and messages with different degrees of registration recommendation may be presented.

  The dictionary correction management unit 21 outputs the dictionary correction registration determination result sent from the dictionary registration determination unit 29 together with the analysis result sent from the dictionary correction influence analysis unit 25 and the language processing result of the processing target sentence sent from the language processing unit 23. Transmit to device 4.

  Specifically, the analysis results of the input sentence 1 and the input sentence 2 are combined with the correction influence score distribution 0.0 ± 0.5 of the input sentence 1 and the correction influence score distribution 0.49 ± 0.43 of the input sentence 2. Further, the determination result that the dictionary should not be registered (an output example is shown in FIG. 36) is transmitted to the output device 4 together with the distribution 0.25 ± 0.53 of the correction influence score in the entire target sentence.

  The output device 4 presents the language processing result, the analysis result, and the dictionary registration determination (step S1402). Specifically, the analysis results of the input sentence 1 and the input sentence 2 are combined with the correction influence score distribution 0.0 ± 0.5 of the input sentence 1 and the correction influence score distribution 0.49 ± 0.43 of the input sentence 2. In addition, a determination result that the dictionary should not be registered (an output example is shown in FIG. 36) is presented together with the distribution 0.25 ± 0.53 of the correction influence score in the entire target sentence.

Next, a fifteenth specific embodiment of the present invention is described.
Here, a specific example of the fifteenth embodiment will be described with reference to FIGS. 16 and 18.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described.

  In the dictionary strengthening support process, the dictionary strengthening support system 150 first executes the processes of step S101 to step S106, step S1401, and step S1402, as in the above-described fourteenth embodiment.

  In presenting the language processing result, analysis result, and dictionary registration judgment (step S1402), the dictionary operator looks at the output result as shown in FIG. 36 and analyzes the “thought” (adverb) that should have been registered in the dictionary. Since it does not appear in the results, it is possible to obtain appropriate analysis results for both input sentence 1 and input sentence 2 by registering not only “thoughts” (adverbs) but also “thinking” (adverbs). Suppose that it is determined.

  When the dictionary worker inputs a dictionary correction registration request from the input device 1, the dictionary addition of “thought” (adverb) is reflected in the master dictionary (steps S1501 and S1502).

  Further, the dictionary correction is continued, and the dictionary correction procedure is restarted from step S101 with the changed master dictionary. When the dictionary operator further inputs a dictionary registration of “Thinking (Adverb)” as a dictionary correction, dictionary correction information is sent from the input device 1 to the dictionary correction management means 21.

  The dictionary correction management means 21 sends a dictionary editing means 22 a temporary registration dictionary generation request for generating a dictionary in which this dictionary correction information is applied to the master dictionary as a temporary registration dictionary. The dictionary editing unit 22 further generates a temporary registration dictionary in which “in thought (adverb)” is added to the master dictionary, and stores the temporary registration dictionary in the temporary registration dictionary storage unit 32 (step S102).

  Next, when the dictionary operator inputs the sentence “each thought” (input sentence 1 in FIG. 20) including the surface layer of the added word “thinking (adverb)” from the input device 1, the dictionary correction management means 21. Sends this input sentence to the processing target sentence selection means 24 as a candidate for the processing target sentence (step S103).

  When the dictionary correction management unit 21 further sends a processing target sentence selection request to the processing target sentence selecting unit 24, the processing target sentence selecting unit 24 includes the surface layer of the word “thought” to be corrected because the input sentence 1 includes A sentence to be processed is selected and sent to the dictionary correction management means 21 (step S104).

  When an analysis request in the input sentence 1 is sent to the dictionary correction influence analysis means 25, a language processing request for inputting the processing target sentence is sent to the language processing means 23. The language processing unit 23 executes language processing for each input sentence based on the master dictionary and the temporary registration dictionary, and obtains a morphological analysis result as shown in FIG. 37 (step S105).

  The dictionary correction influence analysis means 25 obtains the distribution function of the correction influence score of the example sentence 1. Since “decrease in morpheme number” and “generation of corrected (registered) word” in FIG. 35 correspond, the expected value and standard deviation of the synthesized distribution function are μ = 0.49 and σ = 0.43. (Step S106).

  Similarly, the values of the entire target sentence are μ = 0.49 and σ = 0.43. Since μ−σ> 0 is satisfied, it is determined that registration is recommended as a registration determination (step S1401).

  As an evaluation result, an output format as shown in FIG. 38 is presented from the output device 4 (step S1402).

  The dictionary operator looks at the output result as shown in FIG. 38, determines that the input sentence 1 can also obtain an appropriate analysis result by registering “in thought” (adverb), and corrects the dictionary from the input device 1. Is added to the master dictionary, the dictionary addition of “I want to (adverb)” is reflected in the master dictionary (steps S1501 and S1502).

Next, a sixteenth specific embodiment of the present invention will be described.
Here, with reference to FIGS. 16 and 19, a specific example of the above-described sixteenth embodiment will be described.

  In this example, it is assumed that the language processing unit 23 performs morphological analysis processing as language processing. A specific process in the case where the dictionary operator newly adds “thought” (adverb), which is a word to be corrected, to the dictionary will be described.

  In the dictionary strengthening support process, the dictionary strengthening support system 150 first executes the processes of step S101 to step S106 and step S1401 as in the fifteenth embodiment.

  In this example, in the dictionary correction registration determination (step S1401), the distribution of the correction influence score for the input sentence 1 in the entire target sentence is μ = 0.49 and σ = 0.43. Since μ−σ> 0 is satisfied, it is determined that registration is recommended as a registration determination.

  Then, the dictionary correction management means 21 sends a dictionary correction registration request to the dictionary editing means 22. When a registration request for dictionary correction is sent, the dictionary editing unit 22 reflects the addition of the dictionary “done (adverb)” in the master dictionary (steps S1501 and S1502).

  In each of the embodiments described above, the viewpoint of morphological analysis processing that uses Japanese as input has been described as language processing. However, syntax analysis, machine translation, and the like based on morphological analysis processing are also possible. Provides dictionary enhancement support suitable for language processing by performing dictionary modification impact analysis based on comparison of syntax analysis results and machine translation results as analysis results can do.

  According to the present invention, a language processing apparatus that performs language processing based on a word dictionary such as morphological analysis, syntax analysis, and machine translation that provides a dictionary strengthening work support function, and a dictionary strengthening work support function provided to the language processing apparatus. It is useful to apply to the control program to make it. The present invention can also be applied to a Kana-Kanji conversion device, a text search device, and a text summarization device equipped with these language processing functions.

It is a block diagram which shows the structural example of the dictionary reinforcement | strengthening assistance system in the 1st and 7th embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 1st Embodiment of this invention. It is a block diagram which shows the structural example of the dictionary reinforcement | strengthening assistance system in the 2nd, 4th and 8th embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 2nd and 8th embodiment of this invention. It is a block diagram which shows the structural example of the dictionary reinforcement | strengthening assistance system in the 3rd, 5th and 6th embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 3rd Embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 4th Embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 5th Embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 6th Embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 7th Embodiment of this invention. It is a block diagram which shows the structural example of the dictionary reinforcement | strengthening assistance system in the 9th and 10th embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 9th Embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 10th Embodiment of this invention. It is a block diagram which shows the structural example of the dictionary reinforcement | strengthening assistance system in the 11th, 12th and 13th Embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 11th, 12th and 13th Embodiment of this invention. It is a block diagram which shows the structural example of the dictionary reinforcement | strengthening assistance system in the 14th, 15th and 16th Embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 14th Embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 15th Embodiment of this invention. It is a flowchart which shows the example of the dictionary reinforcement | strengthening assistance process in the 16th Embodiment of this invention. It is explanatory drawing which shows the process target sentence in the 1st and 15th Example, and its morphological analysis result. It is explanatory drawing which shows the influence analysis table in 1st Example. It is explanatory drawing which shows the influence analysis result shown in a 1st, 11th, 12th, and 13th Example. It is explanatory drawing which shows the example sentence used in 2nd Example. It is explanatory drawing which shows the example sentence used in 3rd Example. It is explanatory drawing which shows the process target sentence in 6th Example, and its morphological analysis result. It is explanatory drawing which shows the example sentence used in 7th Example. It is explanatory drawing which shows the process target sentence in 7th Example, and its morphological analysis result. It is explanatory drawing which shows the influence analysis result shown in 7th and 10th Example. It is explanatory drawing which shows the example sentence used in the 8th and 10th Example. It is explanatory drawing which shows the example sentence used in 9th Example. It is explanatory drawing which shows the example sentence used in the 9th and 10th Example. It is explanatory drawing which shows the abstract of the influence analysis in 11th Example. It is explanatory drawing which shows the abstract of the influence analysis in 12th Example. It is explanatory drawing which shows the abstract of the influence analysis in 13th Example. It is explanatory drawing which shows the influence analysis table in 14th and 15th Example. It is explanatory drawing which shows the influence analysis result shown in 14th and 15th Example. It is explanatory drawing which shows the process target sentence in 15th Example, and its morphological analysis result. It is explanatory drawing which shows the influence analysis result shown in 15th Example.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Input device 2, 2A, 2C, 2D Data processing device 3, 3A Storage device 21 Dictionary correction management means 22 Dictionary editing means 23 Language processing means 24 Processing target sentence selection means 25 Dictionary correction influence analysis means 26 Example sentence registration means 27 Reference language Processing result registration means 28 Influence analysis abstract creation means 29 Dictionary registration judgment means 31 Master dictionary storage part 32 Temporary registration dictionary storage part 33 Influence analysis table storage part 34 Example sentence storage parts 100, 110, 120, 130, 140, 150 Dictionary strengthening support system

Claims (22)

A dictionary strengthening support system for supporting correction work for strengthening a word dictionary used for natural language processing,
Master dictionary storage means for storing the word dictionary in which word information including words is registered as a master dictionary;
Dictionary editing means for creating a temporary registration dictionary in which the word dictionary is edited based on dictionary editing information indicating correction instruction content of the word dictionary;
Temporary registration dictionary storage means for storing the temporary registration dictionary created by the dictionary editing means;
A processing target sentence selection means for selecting, as a processing target sentence, a sentence including a word that is a correction target in the dictionary editing information, from processing target candidate sentences that are processing target candidates for natural language processing;
Language processing means for executing natural language processing based on each of the master dictionary and the temporary registration dictionary for the processing target sentence selected by the processing target sentence selection means;
The processing result of the natural language processing for the processing target sentence based on the master dictionary by the language processing means is compared with the processing result of the natural language processing for the processing target sentence based on the temporary registration dictionary by the language processing means. A dictionary correction influence analyzing means for analyzing the influence of dictionary editing according to the correction instruction content indicated by the dictionary editing information with respect to a predetermined viewpoint;
A dictionary strengthening support system comprising analysis result presentation means for presenting an analysis result by the dictionary correction influence analysis means together with a processing result of natural language processing. The viewpoint of comparative analysis of the processing result of natural language processing based on the master dictionary and the processing result of natural language processing based on the temporary registration dictionary is set in association with a numerical value indicating the influence of the estimated dictionary editing. An impact analysis table storage means for storing the impact analysis table
The dictionary strengthening support system according to claim 1, wherein the dictionary correction influence analysis means performs a comparative analysis according to the influence analysis table. The viewpoint of comparative analysis is one of the phenomena seen in the difference between the processing result of the natural language processing based on the master dictionary and the processing result of the natural language processing based on the temporary registration dictionary, and is subject to correction. The dictionary strengthening support system according to claim 2, including whether or not a generated word is generated and whether or not the number of morphemes in the processing target sentence is reduced. An example sentence storage means for storing example sentences prepared in advance as processing target candidate sentences;
The dictionary strengthening support system according to any one of claims 1 to 3, wherein the processing target sentence selecting unit selects a processing target sentence from processing target candidate sentences including example sentences stored in the example sentence storage unit. An example sentence registration means for registering an example sentence in the example sentence storage means;
An input sentence receiving means for receiving an input sentence input by a correction worker,
The processing target sentence selecting unit selects a processing target sentence from the processing target candidate sentences including the input sentence received by the input sentence receiving unit,
The dictionary strengthening support system according to claim 4, wherein the example sentence registration unit registers the input sentence received by the input sentence reception unit as an example sentence in the example sentence storage unit. An input request presenting means for presenting an input request for an input sentence as a candidate sentence to be processed to a correction worker;
The processing target sentence selecting means cannot select that there is no processing target candidate sentence that can be selected as the processing target sentence when the sentence including the word that is the correction target in the dictionary editing information is not in the processing target candidate sentence. Send information,
The dictionary strengthening support system according to claim 5, wherein the input request presenting unit presents an input request for the input sentence based on reception of the non-selectable information. The example sentence registration unit registers the input sentence as an example sentence in the example sentence storage unit when the input sentence received by the input sentence reception unit is different from the example sentence registered in the example sentence storage unit. 6. The dictionary strengthening support system according to 6. The processing target sentence selecting unit selects a processing target sentence from processing target candidate sentences including the example sentence stored in the example sentence storage unit and the input sentence received by the input sentence receiving unit,
If the analysis result of the dictionary correction influence analysis means for the input sentence selected as the processing target sentence is different from any of the analysis results of the dictionary correction influence analysis means for the example sentence selected as the processing target sentence, the example sentence registration means The dictionary strengthening support system according to any one of claims 5 to 7, wherein an input sentence is registered in the example sentence storage means as an example sentence. Provided with an input sentence receiving means for receiving an input sentence input by a correction worker,
The input sentence receiving means receives an input of a natural language processing result of the input sentence together with the input sentence,
The processing target sentence selecting means selects an input sentence including a word that is a correction target in the dictionary editing information as a processing target sentence and uses a natural language processing result corresponding to the input sentence selected as the processing target sentence as a reference Select as language processing result,
The dictionary correction influence analysis means includes a processing result of the natural language processing for the processing target sentence based on the master dictionary by the language processing means and a processing result of the natural language processing for the processing target sentence based on the temporary registration dictionary by the language processing means. And the reference language processing result of the processing target sentence selected by the processing target sentence selection means, and analyze the viewpoint in which the influence of dictionary editing according to the correction instruction content indicated by the dictionary editing information is determined in advance And
The analysis result presentation means
The dictionary strengthening support system according to any one of claims 1 to 8, wherein an analysis result by the dictionary correction influence analyzing means is presented together with a processing result of the natural language processing and a reference language processing result. An example sentence storage means for storing example sentences prepared in advance as processing target candidate sentences;
The example sentence storage means stores an example sentence together with a reference language processing result of the example sentence,
The processing target sentence selecting means selects, as a processing target sentence, an example sentence including a word that is to be corrected in the dictionary editing information, and selects a reference language processing result corresponding to the example sentence selected as the processing target sentence. Item 10. The dictionary strengthening support system according to Item 9. An example sentence registration means for registering an example sentence in the example sentence storage means;
A reference language processing result registration means for registering a reference language processing result of an example sentence in the example sentence storage means;
The example sentence registering unit registers the input sentence received by the input sentence receiving unit as an example sentence in the example sentence storage unit,
The reference language processing result registration means registers the natural language processing result received together with the input sentence by the input sentence receiving means as the reference language processing result of the input sentence in the example sentence storage means,
The processing target sentence selecting unit selects a processing target sentence from the processing target candidate sentences including the example sentence stored in the example sentence storage unit and the input sentence received by the input sentence receiving unit, and selects the processing target sentence. The dictionary strengthening support system according to claim 10, wherein a corresponding reference language processing result is selected. A determination result receiving means for receiving a determination result as to whether or not to register the processing result of the natural language processing for the processing target sentence by the language processing means as the reference language processing result;
The example sentence registering unit registers the processing target sentence, which is the input sentence received by the input sentence receiving unit, as the example sentence in the example sentence storing unit in response to receiving the determination result of registration by the determination result receiving unit. And
The reference language processing result registering unit displays the processing result of the natural language processing by the language processing unit for the processing target sentence registered as an example sentence in response to the determination result being registered by the determination result receiving unit being received. The dictionary strengthening support system according to claim 11, wherein the dictionary is registered in the example sentence storage means as a reference language processing result of the example sentence. It has an impact analysis abstract creation means for creating an impact analysis abstract based on the analysis result of the dictionary correction impact analysis means,
The dictionary correction influence analysis means derives a dictionary correction influence score indicating the influence of dictionary editing according to the correction instruction content indicated by the dictionary editing information for each processing target sentence,
The impact analysis abstract creating means creates an impact analysis abstract in which the analysis result and the processing result of natural language processing are arranged based on the dictionary correction impact score,
The dictionary strengthening support system according to any one of claims 1 to 12, wherein the analysis result presentation means presents an abstract of the impact analysis created by the impact analysis abstract creation means. It has an impact analysis abstract creation means for creating an impact analysis abstract based on the analysis result of the dictionary correction impact analysis means,
The impact analysis abstract creation means selects one of a plurality of processing target sentences having the same analysis result, and arranges the analysis result for the processing target sentence having a different analysis result and the processing result of the natural language processing. Create an impact analysis abstract,
The dictionary strengthening support system according to any one of claims 1 to 12, wherein the analysis result presentation means presents an abstract of the impact analysis created by the impact analysis abstract creation means. It has an impact analysis abstract creation means for creating an impact analysis abstract based on the analysis result of the dictionary correction impact analysis means,
The influence analysis abstract creation means determines a processing target sentence to be presented based on an analysis result of the processing target sentence, and analyzes the processing result of the processing target sentence decided to be the presentation target and processing of natural language processing Create an impact analysis abstract alongside the results,
The dictionary strengthening support system according to any one of claims 1 to 12, wherein the analysis result presentation means presents an abstract of the impact analysis created by the impact analysis abstract creation means. Based on the analysis result of the dictionary correction influence analysis means, it is provided with a dictionary registration determination means for determining whether or not dictionary editing according to the correction instruction content indicated by the dictionary editing information is reflected in the master dictionary,
The dictionary strengthening support system according to any one of claims 1 to 15, wherein the analysis result presenting unit presents a determination result of the dictionary registration determining unit. A dictionary registration determination receiving means for receiving a determination result as to whether or not to reflect dictionary editing in accordance with the correction instruction content indicated by the dictionary editing information in the master dictionary;
The dictionary editing means according to claim 16, wherein the dictionary editing means reflects dictionary editing in accordance with the correction instruction content indicated by the dictionary editing information in the master dictionary in accordance with the determination result accepted by the dictionary registration determination accepting means. Support system. A dictionary strengthening support method for supporting a correction work for strengthening a word dictionary in which word information including a word, which is a master dictionary used for natural language processing, is registered,
Creating a temporary registration dictionary obtained by editing the word dictionary based on dictionary editing information indicating correction instruction content of the word dictionary;
Selecting, as a processing target sentence, a sentence including a word that is a correction target in the dictionary editing information from a processing target candidate sentence that is a processing target candidate for natural language processing;
Performing natural language processing based on each of the master dictionary and the temporary registration dictionary for the selected processing target sentence;
The correction instruction content indicated by the dictionary editing information by comparing the processing result of the natural language processing for the processing target sentence based on the master dictionary and the processing result of the natural language processing for the processing target sentence based on the temporary registration dictionary Analyzing the impact of dictionary editing according to a predetermined point of view;
A method for supporting dictionary strengthening, comprising: performing an analysis result along with a processing result of natural language processing. 19. The dictionary strengthening support method according to claim 18, wherein a step of selecting a processing target sentence from processing target candidate sentences including example sentences stored in an example sentence storage unit that stores example sentences prepared in advance as processing target candidate sentences is executed. Receiving the input of the natural language processing result of the input sentence together with the input sentence input by the correction worker;
A step of selecting an input sentence including a word to be corrected in the dictionary editing information as a processing target sentence and selecting a natural language processing result corresponding to the input sentence selected as the processing target sentence as a reference language processing result When,
The processing result of the natural language processing for the processing target sentence based on the master dictionary, the processing result of the natural language processing for the processing target sentence based on the temporary registration dictionary, and the reference language processing result of the selected processing target sentence. Comparing and analyzing the influence of dictionary editing according to the correction instruction content indicated by the dictionary editing information on a predetermined viewpoint;
The dictionary strengthening support method according to claim 18 or 19, wherein an analysis result is presented together with a processing result of the natural language processing and a reference language processing result. Deriving a dictionary correction influence score indicating the influence of dictionary editing according to the correction instruction content indicated by the dictionary editing information for each processing target sentence;
Creating an impact analysis abstract in which the analysis results and the processing results of natural language processing are arranged based on the dictionary-corrected impact score;
The dictionary strengthening support method according to any one of claims 18 to 20, wherein an abstract of impact analysis is presented. A dictionary strengthening support program for supporting a correction work for strengthening a word dictionary in which word information including a word, which is a master dictionary used for natural language processing, is registered,
On the computer,
Creating a temporary registration dictionary obtained by editing the word dictionary based on dictionary editing information indicating correction instruction content of the word dictionary;
Selecting, as a processing target sentence, a sentence including a word that is a correction target in the dictionary editing information from a processing target candidate sentence that is a processing target candidate for natural language processing;
Performing natural language processing based on each of the master dictionary and the temporary registration dictionary for the selected processing target sentence;
The correction instruction content indicated by the dictionary editing information by comparing the processing result of the natural language processing for the processing target sentence based on the master dictionary and the processing result of the natural language processing for the processing target sentence based on the temporary registration dictionary Analyzing the impact of dictionary editing according to a predetermined point of view;
A dictionary strengthening support program for executing the step of presenting the analysis result together with the processing result of the natural language processing.

Images