JP6018932B2 - Example search device, example search method, and example search program - Google Patents

Description

  The present invention relates to an example search technique for supporting document creation.

  In recent years, offshore development has increased. In offshore development, a local engineer whose Japanese is a non-native language may create a design document, etc., and a non-native speaker is required to support design document creation in Japanese. When a non-native speaker who is not familiar with Japanese creates a design document in Japanese, create a document that contains grammatical errors (especially errors in the use of “particles” specific to Japanese). This leads to a reduction in the quality of design documents. One example of support for non-native speakers who are not well-versed in Japanese grammar is to create documents that are grammatically correct.

  Conventional example search techniques include a technique for searching examples using text as an input and a technique for searching examples using words as input. As an example of the former (searching with text as input), for example, there is a technique described in Patent Document 1. In the technique described in Patent Document 1, an input sentence (search keyword sentence) is case-analyzed, and a sentence including two independent words in the search keyword sentence and having the same case relation is preferentially displayed as a result. However, in this technique, it is necessary to input grammatically correct Japanese at the time of searching in order to collate the case relationship between the search keyword sentence and the sentence in the search document to be searched. If a search keyword sentence having a grammatical error is input, an example containing the grammatical error is preferentially displayed. Therefore, when a user who is a non-native speaker who is difficult to input grammatically correct Japanese in the first place uses the search technology described in Patent Document 1, it is difficult to search for correct Japanese after all. There's a problem.

  On the other hand, in the latter case (searching using words as input), even a user who is not familiar with Japanese grammar can search for a correct example, but an example sentence that shows a usage that does not conform to the user's intention ( Hereinafter, “example” or “example sentence” (the same applies throughout the specification) is included in the search result, which increases the burden on the user when selecting a desired example sentence from the search result. is there. Specifically, for example, if the user has doubts about whether to “transition screen” or “transition to screen” should be adopted, first, “screen” and “ Search by the word “transition”. In the result, for example, an example sentence that the user does not want to use is output, such as “screen transition”, and the user must select an example sentence to be adopted from the search result.

  As a technique for reducing such a burden on the user, there is a technique disclosed in Patent Document 2, for example. In this technique, a storage unit storing a case pattern is stored in advance, and a case pattern including a search key (or “search keyword” in some cases; the same applies throughout the specification) is used with an input word as a key. Displayed to the user and prompts for selection. When the user specifies an appropriate case pattern for the displayed case pattern, the search system searches the text database for example sentences that match the specified case pattern, and presents them to the user. .

Japanese Patent Laid-Open No. 7-244669 JP-A-9-81588

  A word-based search technique is effective for users who are not familiar with Japanese grammar to search Japanese example sentences. However, in order to classify and display the search results so that the user can easily use them, the search system side holds a case pattern (semantic feature) for each word as in the technique disclosed in Patent Document 2. It is necessary to keep. However, the work of creating a dictionary with a case pattern is actually performed by humans, which takes a lot of labor and effort. Furthermore, if there is a word without a case pattern in the example sentence to be searched, that example sentence cannot be processed. That is, such example sentences cannot be classified and cannot be included in the result display. If the example sentences are displayed separately from the classified example sentences, there is a problem in that the display contents are miscellaneous and the user has to extract troublesome examples from the search results. Furthermore, since it is necessary for the user to select a case pattern at the time of the search, when a user who does not understand the grammar correctly uses this search technology, there arises a problem that it is difficult to obtain an intended search result. .

  The various problems described above are problems that can occur in example searches not only when creating a design document but also when creating all documents.

  Accordingly, an object of the present invention is to provide an example search technique that can more easily search for an example that matches the user's intention in order to solve the above-described problems.

  In order to solve the above problems, an embodiment of the present invention includes a sample sentence, a morpheme analysis result that is a result of automatically performing morphological analysis on the sample sentence, and a structure analysis result that is a result of automatically performing syntactic structure analysis. Are stored in the example storage unit that includes a word that has a combination of part of speech and a word that is the same combination as the input word. The example search apparatus includes a processing unit that classifies the input example sentences based on the nouns, particles, verbs including the input words, and patterns of their use.

  In another aspect of the present invention, an example sentence, a morpheme analysis result that is a result of automatic morphological analysis of the example sentence and a structure analysis result that is a result of automatic syntactic structure analysis are stored in association with each other. An example search method executed by a computer device including an example storage unit, a part of speech giving unit, and a processing unit, wherein the part of speech giving unit gives a part of speech to an input word, and the processing unit Example sentences stored in the example storage unit including words whose word-of-speech and word correspondence are the same combination as the input word are used as nouns, particles, verbs including the input word, and patterns of their use. And a step of classifying based on the example search method.

  In another aspect of the present invention, an example sentence, a morpheme analysis result that is a result of automatic morphological analysis of the example sentence and a structure analysis result that is a result of automatic syntactic structure analysis are stored in association with each other. A computer device comprising an example storage unit, a part-of-speech providing unit, and a processing unit, wherein the part-of-speech providing unit assigns a part-of-speech to an input word; Classifying example sentences stored in the example storage unit including words having the same combination as the input words based on nouns, particles, verbs including the input words, and patterns of their use. It is an example search program for making it happen.

  According to the present invention, it is possible to easily retrieve an example that matches the user's intention.

It is a figure which shows an example of a structure of the example search device which concerns on one Embodiment of this invention. It is a main flowchart which shows an example of the flow of the whole process of the example search device which concerns on one Embodiment of this invention. It is a flowchart which shows an example of the flow of a process in the part of speech provision part of the example search apparatus which concerns on one Embodiment of this invention. It is a figure which shows an example of the data for determining the part of speech provided to the morpheme of a keyword. It is a figure which shows an example of the data for determining the part of speech provided to a keyword. It is a flowchart which shows an example of the flow of a process in the example sentence extraction part of the example search device which concerns on one Embodiment of this invention. It is a flowchart which shows an example of the flow of a process in the example sentence classification | category part of the example search device which concerns on one Embodiment of this invention. It is a flowchart which shows an example of the flow of a process in the example sentence rearrangement part of the example search device which concerns on one Embodiment of this invention. It is a figure which shows an example of the data which show the priority of the part of speech of the example sentence which should be preferentially shown according to the part of speech of a search keyword. It is a flowchart which shows an example of the flow of a process in the result output part of the example search device which concerns on one Embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described. The example search apparatus according to the present embodiment may include words in various languages (phrases, sentences, etc. The same applies hereinafter. In the present embodiment, the case of Japanese will be described as an example). An example sentence including a key word is extracted from the example database (DB) as a search key. Then, the extracted example sentences are classified for each or all combinations of “noun”, “particle”, “verb”, and “utilization (inflection)” including the search key, and the user (hereinafter referred to as “user”). The representative sentence to be displayed as a search result is selected, and the selected representative sentence is displayed and output on an output device such as a display. The feature of the example search device according to the present embodiment is that, after extracting an example sentence using a word as a key, the part of speech of the search key in the extracted example sentence is considered without using case information for each word. . Japanese is basically composed of “noun”, “particle”, “verb”, and “utilization”, and the example search device of this embodiment includes a word that is the same as the search key and has the same part of speech, And a combination of some or all of the “noun”, “particle”, “verb”, “utilization” (hereinafter referred to as “quadruple”) that are in a dependency relationship with the word The extraction results are classified for each combination). If the user wants to understand examples of how to use these quadruples, there is no need to classify the example sentences by case patterns, but by classifying the example sentences by focusing on the role of part of speech, Present.

  Hereinafter, an example search device according to the present embodiment will be described in detail with reference to the drawings. In the drawings referred to in the following description, the same parts as those in the other drawings are denoted by the same reference numerals.

(Configuration of example search device)
FIG. 1 is a diagram illustrating an example of a configuration of an example search apparatus according to the present embodiment. The example search device 1 includes an input / output unit 102, a part-of-speech adding unit 104, a sample sentence extracting unit 106, a sample sentence classification unit 108, a sample sentence rearranging unit 110, a result output unit 112, and a sample storage unit 114.

  The input / output unit 102 accepts a search keyword (word / phrase) input by a user using an input device. Further, the search result of the example sentence output by the result output unit 112 is output to the output device 30 such as a monitor.

  The part-of-speech giving unit 104 gives part-of-speech to a search keyword input by a user input operation. In the present embodiment, the example search device 1 further includes a part-of-speech information storage unit 116 that stores a correspondence relationship between a morpheme condition and a part-of-speech to be assigned to an input word. The part of speech is given to the word input with reference to the part of speech information storage unit 116.

  The example sentence extraction unit 106 extracts, from the example storage unit 114, an example sentence including a search keyword and syntax structure information that is a morphological analysis result and a structure analysis result of the example sentence. The example storage unit 114 stores an example sentence, a morphological analysis result of the example sentence, and a structure analysis result of the example sentence in association with each other.

  The example sentence classification unit 108 creates a combination of nouns, particles, verbs, and some or all of the utilizations for the search keywords included in the example sentences, and classifies the example sentences into example groups for each combination.

  The example sentence rearranging unit 110 groups the example groups into large groups according to nouns, particles, verbs, or usage priority determined according to the part of speech of the search keyword.

  The result output unit 112 outputs an example sentence of the search result.

  The example sentence extraction unit 106 and the example sentence classification unit 108 correspond to the “processing unit” in the present invention.

  The example search device 1 can be realized by a configuration similar to that of a general computer such as a CPU (Central Processing Unit), a memory such as a RAM, a storage device such as a hard disk, a network interface, etc. (not shown). . The function of each component of the example search device 1 is, for example, when the CPU reads and executes a program stored in a hard disk or the like, or, for example, a custom design of sequencer logic in an FPGA (Field Programmable Gate Array) It is a function that is realized. The data used in the processing in each configuration is data stored in the hard disk or RAM of the example search device 1.

(Processing flow of example search method)
Hereinafter, a processing flow in the example search device according to the present embodiment will be described with reference to a flowchart.

(Main flow)
FIG. 2 is a main flow diagram showing an example of a rough flow of overall processing in the example search device of the present embodiment.

  When the user searches for an example sentence, when the keyword 10 to be searched is input by operating the input device 20 such as a keyboard, the input / output unit 102 stores one or more keywords 10 input by the user. Accept input.

  The part-of-speech giving unit 104 gives an appropriate part-of-speech to the input keyword (step S10). In parallel with or after this, the example sentence extraction unit 106 searches the example storage unit 114 using the input keyword as a key, and extracts an example sentence that includes the keyword (step S20).

  The example sentence classifying unit 108 creates a set of four combinations in the search keyword for the example sentence extracted in step S20, and classifies each example sentence into an example group based on the created combination of the four sets (step S30). Then, the example sentence rearranging unit 110 classifies the example group into a large group according to the part of speech of the search keyword input by the user. Then, the example groups are rearranged in descending order of the number of examples in the large group, and the example sentences are rearranged in the example groups included in each large group (step S40). Finally, an example sentence is extracted for each rearranged example group, and displayed as an example search result to the user (step S50).

Hereinafter, the details of the processing in each step of Steps S10 to S50 will be described using a flowchart.
(Processing flow of step S10: Processing of part of speech giving unit)
Next, details of the processing in step S10 of FIG. 2 will be described. Step S10 in FIG. 2 is a process executed by the part of speech providing unit 104. FIG. 3 is a flowchart showing details of the processing in step S10 of FIG.

  The part-of-speech providing unit 104 receives an input list (referred to as input list A) including one or more keywords 10 input by the user from the input / output unit 102, and extracts one keyword from the input list A (step) S101). A keyword may be a word, a phrase, or a sentence. Further, as a delimiter between words, a predetermined symbol such as a comma, a punctuation mark or a space may be used. The embodiment shows an example in which a space is used as a break between words. In the following description, the “input list” means a set of one or more input data received by each component of the example search device 1 described below.

  A morphological analysis is performed on the keywords extracted in step S101 using an existing morphological analysis engine (step S102). One morpheme is extracted from the result of morpheme analysis (step S103).

  Whether the morpheme acquired in step S103 matches any of the conditions in Table 1 shown in FIG. 4 is determined in order from the higher condition (record) in Table 1 (step S104). Here, Table 1 is data indicating a part-of-speech condition to be given to a keyword morpheme extracted by morphological analysis, and is data stored in the part-of-speech information storage unit 116, for example. Specifically, the data in Table 1 is stored in a storage device such as a hard disk of the example search device 1 in a format such as a database.

  Because words may be used differently depending on the nature of the target document, the part of speech assigned to the search keyword is changed using the part of speech information storage unit that matches the nature of the target document (part of speech given by morphological analysis) By adding a process that corrects), it is possible to present more useful example sentences at the time of example search (the part of speech given in the morphological analysis may be used as it is).

  As a reason for using the part-of-speech information storage unit 116, for example, in the case of a design document, the expression of a body stop is often used. As a specific example, the word “display” is generally used in the form of “display XX” rather than “display XX”. When morphological analysis is performed on the expression “display XX”, the part of speech given to “display” becomes a noun. Rather than extracting the quaternary set described later with the displayed part of speech as "noun", the displayed POS is regarded as a "verb" and focusing on the object (noun) and display (verb) Can use this technique to appropriately extract the particle pattern between the object “XX (noun)” and the “indication” that is the body stop. Therefore, when the document to be processed is a design document, a rule or word to be given a part of speech other than the part of speech assignment result in morphological analysis is stored in advance in the part of speech information storage unit 116, and the result of simple morphological analysis The ability to extract effective example sentences can be improved by adding a configuration that allows a part of speech different from that to be given to a search keyword.

  Returning to FIG. 3, if there is no condition that matches Table 1, the part of speech obtained by the morphological analysis in step S103 is directly applied to the morpheme to be processed (step S105). If there is a matching condition, the part of speech corresponding to the matching condition in Table 1 is assigned to the morpheme to be processed (step S106).

  When there is an unprocessed morpheme among the morphemes obtained in step S102 (step S107: No), the process returns to step S103. If all the morphemes have been processed (step S107: Yes), whether or not the part-of-speech of the keyword to be processed matches the conditions in Table 2 shown in FIG. The determination is made in order (step S108). Here, Table 2 is data indicating a part-of-speech condition to be given to a keyword, and is data stored in the part-of-speech information storage unit 116, for example. Specifically, the data in Table 2 is stored in a storage device such as a hard disk of the example search device 1 in a format such as a database.

  If there is no condition that matches Table 2, the part of speech given to the morpheme in the processing of steps S103 to S107 is given to the keyword to be processed as it is. At this time, if the keyword is a phrase or a sentence, the part of speech for each word is concatenated with “+” (step S109). If there is a matching condition, the part of speech corresponding to the matching condition in Table 2 is assigned to the keyword to be processed. At this time, if the keyword is a phrase or a sentence, the part of speech is determined with reference to Table 2 for each word included in the keyword, and these parts of speech are concatenated with “+” (step S110).

  Finally, it is determined whether or not there is an unprocessed keyword in the input list A (step S111). If there is an unprocessed keyword, the process returns to step S101. When all the keywords have been processed, the input example list A to which the part-of-speech information is assigned is transferred to the example sentence classifying unit 108, and the process ends.

  Here, as a method for determining whether or not there is an unprocessed keyword, for example, when the input list A is received, the total number of keywords in the list is obtained based on delimiter information (space in this example). (For example, if there are N spaces in the list, the number of words is N + 1). In addition, a counter (initial value = 0) for storing the input number (the number of processed keywords) is provided. In step S101, the “number of counters + 1” -th keyword is extracted in order from one end of the input list A, and after the keyword is extracted, the number of counters is incremented by one. In step S103, if the number of counters is less than the total number of keywords in the input list A, it is determined that there are unprocessed keywords.

  For example, if the part-of-speech assigning unit 104 accepts a keyword “screen display” (with a space as a word break) as the input list A, each part of speech is displayed as “screen (noun) display (verb)”. Is granted. As another example, if the keyword “display on the screen” (example using a space as a word break) is accepted as the input list A, the condition is “display (noun + particle) (verb)”. Is given a part of speech.

(Processing flow of step S20: processing of example sentence extracting unit)
Next, details of the processing in step S20 of FIG. 2 will be described. Step S20 in FIG. 2 is a process executed by the example sentence extracting unit 106. FIG. 6 is a flowchart showing details of the process in step S20 of FIG.

  Similar to the part-of-speech providing unit 104, the example sentence extracting unit 106 receives from the input / output unit 102 an input list A including one or more keywords 10 input by the user. A keyword may be a word, a phrase, or a sentence. In this example, a space is used as a break between words. Then, the example storage unit 114 is searched for example sentences including all the keywords included in the input list A (step S201).

  Here, the example storage unit 114 stores an example sentence and additional information (morpheme analysis result and syntax structure information) in association with each other. More specifically, it is a storage area provided in a storage device such as a hard disk that holds data in a data format such as a database. In addition, the data stored in the example storage unit 114 was parsed using a result of morphological analysis of a Japanese document or the like (grammatically correct) using an existing morphological analysis engine and syntactic structure analysis. The result may be automatically registered with the part of speech given based on the rules of Table 1 (FIG. 4) and Table 2 (FIG. 5). By automatically registering the result automatically analyzed by the existing morphological analysis and syntax structure analysis engine, it is possible to prepare data without taking time and effort. Further, when preparing data in the example storage unit 114 in advance, it is more preferable to perform data registration by performing morphological analysis on a document of the same type as the document to be created by the user of the example search apparatus 1. For example, when a user intends to create a detailed design document based on a basic design document (described by a Japanese native), the basic design document and its accompanying documents are analyzed in advance by morphological analysis and syntactic structure analysis. By storing the example in the example storage unit 114, it is highly possible that an example sentence that matches the user's wish is extracted because the target fields are the same.

  Returning to FIG. 6, when the corresponding example sentence does not exist in the example storage unit 114 (step S202: No), a message indicating that there is no corresponding example sentence on the screen viewed by the user is displayed (step S204). . If there is a corresponding example sentence (step S202: Yes), the example sentence and additional information (morpheme analysis result and syntax structure information) are acquired from the search result (step S203). After the acquisition, the example sentence and the additional information are transferred to the example sentence classification unit 108, and the process is terminated.

Specific examples are shown below. For example, it is assumed that the example sentence extracting unit 106 receives “screen display” as the keyword 10 as in the above example. The example sentence extraction unit 106 searches the example storage unit 114 to extract example sentences including these keywords. For example, assume that the following example sentences are extracted.
Example sentence: “Display the following information on the display screen that the user can see”

  Further, the example sentence extracting unit 106 acquires the following additional information associated with the example sentence in the example storage unit 114.

(Form of additional information)
Information on the word, part of speech, start position, and number of characters of each phrase is given in the dependency information. That is, it is as follows. (If there is word information, the underscore “_” is connected as a delimiter.)
(Relationship source) clause number: Word information: Part of speech information: Start position: Number of characters => (Relationship destination) clause number: Word information: Part of speech information: Start position: Number of characters

(Additional information)
0: Below-+: Noun + particle: 1-3 => 1: Information-+: Noun + particle: 4-3 1: Information-+: Noun + particle: 4-3 => 3: Visual Yes: Verb-basic form-Yes: 11-4 2: User-+: Noun + particle: 7-3: Visibility-To do: Verb-Basic form-To: 11-4 3: Visibility-Yes: Verb-Basic form-To : 11-4 => 4: display-screen- +: noun + particle: 15-9 4: display-screen- +: noun + particle: 15-9 => 5: display-verify: verb-basic form- Yes: 24-4

  In this example of additional information, a prepared example sentence is subjected to morphological analysis and syntax structure analysis, and a part of speech, a start position, and a character length are given to each word or phrase. Furthermore, a pair of words or phrases having a dependency relationship is connected by “=>”. Note that the additional information in this example is merely an example, and other expression methods may be adopted.

(Processing flow of step S30: processing of example sentence classification unit)
Next, details of the process in step S30 of FIG. 2 will be described. Step S30 in FIG. 2 is a process executed by the example sentence classification unit 108. FIG. 7 is a flowchart showing details of the process in step S30 of FIG.

  The example sentence classification unit 108 includes one or more keywords 10, part of speech of each keyword obtained by the part of speech adding unit 104, an example sentence obtained by the example sentence extracting unit 106 and its additional information (input list C and Accept). Then, one example sentence is extracted from the input list C (step S301). Next, the word information of the additional information of the extracted example sentence and the keyword are collated (step S302). That is, it is determined whether a keyword is included in the word / phrase of the additional information. If not included (step S303: No), the process returns to step S301 (step S303).

When the keyword is included in the word information of the additional information (step S303: Yes), the part of speech of the word information of the additional information obtained in step S303 is compared with the part of speech of the search keyword (step S304). It is determined whether the part of speech of the keyword and the part of speech of the word information in the additional information obtained in step S303 are the same (step S305). If the parts of speech are not the same, the example group “others” is classified (step S306), and the process returns to step S301.
If the part of speech is the same, create all (partially or all) combinations of the four pairs (“noun”, “particle”, “verb”, “utilization”) for this keyword, including the keyword of interest ( Specifically, a combination of four is extracted) (step S307). The combination of the four sets can be extracted from the dependency information of the additional information and the part of speech information.

For example, “screen display” is accepted as a keyword, and “screen” is a noun, “display”
A case where the part of speech of the verb is given will be described as an example. The “part or all (combination)” mentioned above is only a part of a keyword, that is, (A) “screen (noun)” only,
(B) In the case of “display (verb)” only,
And if you have all the keywords, that is, (C) both “screen (noun)” and “display (verb)”,
Indicates.

  Further, “create all combinations” described above indicates that four sets are extracted for each of (A), (B), and (C) in the above example. For example, for (A), “screen (noun)” is used as a clue, and “participant” associated with this “screen (noun)” and “screen (noun)” in the example sentence are related to “screen (noun)”. “Verb” and its “utilization” are extracted and combined with “screen (noun)” into a set of four.

  Also for (B), using “indication (verb)” as a clue, extract “noun” and its “participant” that are related to “use” and “indication (verb)” into four sets .

For (C), the dependency relationship of “screen (noun)” is “display (verb)”, and the “particle” and “display (verb) associated with“ screen (noun) ”are extracted. The “utilization” of “” is extracted and combined with “screen (noun)” and “display (verb)” to form a set of four. When there are two or more keywords, (C) is acquired preferentially. (If there is one keyword, (A) or (B) will be implemented according to the part of speech of the keyword. If there are two keywords, (C) will be implemented. (Sequentially execute (C) for all combinations selected.)
Returning to FIG. 7, in the processing after step S308, processing for creating an example group for classifying example sentences for each of the same four combinations created in step S307 is performed.

  First, it is determined whether the combination of the four sets created in step S307 already exists in the example group (that is, whether an example group for classifying the combination of the four sets has been created) (step S308). . If there is no example group for classifying the combination of the four sets, this example group is newly created, and the example sentence to be processed is stored in the created example group (step S309). If an example group for classifying the combination of the four sets already exists, the example sentence to be processed is stored in the existing example group (step S310).

  It is determined whether there is an unprocessed example sentence in the input list C (step S311). If there is an unprocessed example sentence, the process returns to step S301. If all the example sentences have been processed, the classification result by this flow is transferred to the example sentence rearranging unit 110, and the process is terminated.

  Hereinafter, a specific example of processing in the example sentence classification unit 108 will be described following the specific example described in the explanation of the example sentence extraction unit 106. For example, it is assumed that the example sentence classification unit 108 receives an input list C having the following contents.

[Input list C]
Keywords (parts of speech): "screen (noun)""display(verb)"
Example sentences (5 search results):

Example sentences (1)
(Example sentence) Display the electronic approval screen on a separate screen (additional information)
0: e-payment-screen- +: e-payment screen: noun + particle: 1-7 => 2: display-display: display: verb-basic-form: 12-4 1: separate-screen- + In: In another screen: Noun + particle: 8-4 => 2: Display-display: Display: verb-basic-select: 12-4

Example sentences (2)
(Example sentence) Set the button type to be displayed on the next screen from the screen ID (additional information)
0: From screen-ID- +: From screen ID: Noun + particle: 1-6 => 4: Setting-setting: Setting: Verb-basic form-: 23-4 1: Next-screen- +: Next On the screen: noun + particle: 8-4 => 2: display-do: display: verb-basic form-do: 12-4 2: display-do: display: verb-basic form-do: 12-4 => 3: button-type- +: button type: noun + particle: 16-7 3: button-type- +: button type: noun + particle: 16-7 => 4: set-set: set : Verb-basic form-to: 23-4

Example sentences (3)
(Example sentence) Display the following information on the screen (additional information)
0: below- +: below: noun + particle: 1-3 => 1: info- +: info: noun + particle: 4-3 1: info- +: info: noun + particle: 4-3 => 3: display-display: display: verb-basic form-select: 10-4 2: screen- +: on screen: noun + particle: 7-3 => 3: display-display: display : Verb-Basic form-To: 10-4

Example sentences (4)
(Example sentence) The work pattern entered in the schedule is displayed on the calendar screen (additional information)
0: Calendar-screen- +: On calendar screen: Noun + particle: 1-8 => 4: Display: Display: Verb-sa modification-Display: 23-2 1: Schedule- + To: Schedule: Noun + Particles: 9-3 => 2: Input-do-ta: Entered: verb-consecutive_basic-do: 12-4 2: Input-do-ta: Entered: verb-intermediate_basic-do: 12- 4 => 3: work-pattern- +: work pattern: noun + particle: 16-7 3: work-pattern- +: work pattern: noun + particle: 16-7 => 4: display: display : Verb-Sa deformation-Indication: 23-2

Example sentences (5)
(Example sentence) Click here to display the decision document screen (additional information)
0: Click: Click: Noun: 1-4_To: To: Verb: 5-2_ To: To: Particle: 7-1 => 2: Display-to-re: Displayed: Verb-Nell-rel connection _Basic form-To be: 15-5 1: Final decision-Document-Screen-+: Final decision document screen: Noun + particle: 8-7 => 2: Display-To be displayed: Displayed: Verb-Unreal Connection_Basic form-Done: 15-5

Then, in step S307 (when passing through steps S302 to S305), the following four combinations are created for each example sentence. At this time, all four different combinations for the search keyword are extracted in each example sentence. For example, in the example sentence (1), for the search keywords “screen (noun)” and “display (verb)”, two different quadruples ([display on, display], [display, display, Yes], both are extracted.
(1) (i) [Display and display on screen] (ii) [Display and display on screen]
(2) [Display on screen]
(3) [Display on screen]
(4) [Display on screen,-]
(5) [Screen is displayed and displayed]

In step S310, the example sentences are classified into example groups as follows. (In addition, when the combination of these four groups already exists in the example group (the combination of the four groups already appears), it is stored in the existing group.)
(A) [display on screen, display] ... Example sentence (1-i)
(B) [Display on screen, display] ... Example sentence (2), Example sentence (3)
(C) [screen, display,-] ... Example sentence (4)
(D) [screen, displayed,] example sentence (5)
(E) [display, display screen] .... Example sentence (1-ii)
(In this example, the combination of the four sets in [] is a group ID for identifying each example group.)

(Processing flow of step S40: processing of example sentence rearranging unit)
Next, details of the processing in step S40 of FIG. 2 will be described. Step S40 in FIG. 2 is a process executed by the example sentence rearranging unit 110. FIG. 8 is a flowchart showing details of the process in step S40 of FIG.

  Accepts an input list (referred to as input list D) including one or more keywords 10, part of speech of each keyword obtained by the part of speech providing unit 104, and an example group obtained by the example sentence classification unit 108 from the example sentence classification unit 108. . Then, Table 3 shown in FIG. 9 is compared with the part of speech of the keyword, and the part of speech for each priority order is acquired (step S401). Here, Table 3 is data indicating the priority order of part of speech of an example sentence to be preferentially selected according to the part of speech of the search keyword. The data in Table 3 is determined in advance based on the results of analyzing the tendency of grammatical errors in Japanese from documents written by non-native speakers who are not familiar with Japanese grammar. It is data. The data in Table 3 is specifically stored in a storage device such as a hard disk of the example search device 1 in the form of a database or the like.

  Next, a group is created for each part-of-speech word with the highest priority for the combination of four sets of group IDs. In this example, for convenience of explanation, this group is referred to as a “large group” (step S402). Next, the number of groups of each large group is calculated (step S403). Rearrange from the large group with the largest number of groups (step S404). As for the rearranged result, one large group is extracted (in order from the top) (step S405).

  Next, one example group is extracted from the extracted large group (step S406). The number of examples in the extracted example group is calculated (step S407). Further, one example sentence is extracted from the example group (step S408). Next, the number of characters in the sentence is calculated as the sentence length of the example sentence, and the example sentence and the sentence length are associated with each other and stored in a storage device such as a hard disk (step S409).

  Using the number of examples calculated in step S407, it is determined whether or not there is an unprocessed example sentence in the target example group (step S410). If there is an unprocessed example sentence, the process returns to step S407. If all the example sentences in the target example group have been processed, the example sentences are rearranged in the example group in ascending order of sentence length (step S411). Here, in this example, the example sentences are rearranged in order from the shortest sentence length. However, when a user searches for an example, it is more grammatically to use a short example sentence than a long example sentence. Even when the example sentence output from the example search apparatus 1 is applied in a simple and easy-to-understand manner, there are merits that a short example sentence is more easily deformed.

  It is determined whether there is an unprocessed example group from the large group (step S412), and if there is, the process returns to step S406. If all the example groups in the target large group have been processed, the example groups are rearranged in descending order of the number of examples in the large group (step S413). Here, in this example, the example groups are rearranged in descending order of the number of examples. The example sentence is extracted from the example storage unit 114, and the example sentence stored in the example storage unit 114 stores a grammatically correct example sentence in Japanese. In addition, if a sentence is extracted from a Japanese document that serves as an example and is automatically registered through morphological analysis and syntactic structure analysis, the large number of examples means that the sentence is used frequently. This means that there is a higher possibility that the example sentence is desired by the user.

  It is determined whether there is an unprocessed large group (step S414). If there is an unprocessed large group, the process returns to step S405. If not, the rearranged result is transferred to the result output unit 112, and the process ends.

  Hereinafter, a specific example of processing in the example sentence rearranging unit 110 will be described following the specific example described in the explanation of the example sentence classification unit 108. For example, it is assumed that the example sentence rearranging unit 110 receives output data (input list D) having the following content from the example sentence classification unit 108.

[Input list D]
Keywords (parts of speech): "screen (noun)""display(verb)"
(A) [display on screen, display] ... Example sentence (1-i)
(B) [Display on screen, display] ... Example sentence (2), Example sentence (3)
(C) [screen, display,-] ... Example sentence (4)
(D) [screen, displayed,] example sentence (5)
(E) [display, display screen] .... Example sentence (1-ii)

(The contents of each example sentence are as follows.)
Example sentence (1) Example sentence for displaying the electronic decision screen on a separate screen (2) Example sentence for setting the button type to be displayed on the next screen from the screen ID (3) Example sentence for displaying the following information on the screen (4) ) On the calendar screen, click the work pattern entered in the schedule and click on the display example sentence (5) to display the final approval document screen

In this case, the following processing is executed in step S401. That is, referring to Table 3, the priority order corresponding to the part of speech (noun, verb) of the keyword in the input list D is as follows.
1st: Particles 2nd: Utilization.

Therefore, in step S402, the example group is classified by “particle” after “screen”. That is, in this example, the following large group is created.
Large group:
(1) [Display and display on screen], [Display and display on screen,-] ... (group of particles = "de")
(2) [Displayed on the screen] (... Particles = "Ni" group)
(3) [Screen, is displayed, ...] (Group of particles = "ga")
(4) [Display, display, display] ... (Particle = "Group")

  Since the large group (1) has two groups and the large groups (2), (3), and (4) each have one group (step S403), the large groups (1) and (2) ), (3), and (4), the number of groups increases in this order (step S404).

  Then, by the processing from step S405 to step S414, in each large group, the example groups classified into the large group are rearranged in descending order of the number of examples. Further, in each example group, the length of the example sentence The example sentences are sorted in ascending order.

Through the processing as described above, the following result is obtained as the processing rearrangement result in the example sentence rearranging unit 110.
1. [Display on screen, display] ... Example sentence (1-i)
2. [Screen, display,-] ... Example sentence (4)
3. [Display on screen, display] ... Example sentence (3), Example sentence (2)
4). [Screen is displayed] ... Example sentence (5)
5. [Display, display, screen] ... Example sentence (1-ii)

  In the above example, a large group is set based on the part of speech with the first priority in Table 3. However, in the large group, the part of speech with the second highest priority (in the above example, “ A group subordinate to the large group may be provided based on “utilization” after “display”. The same applies when the priority is 3rd or lower.

(Processing flow of step S50: processing of result output unit)
Next, details of the processing in step S50 of FIG. 2 will be described. Step S50 in FIG. 2 is a process executed by the result output unit 112. FIG. 10 is a flowchart showing details of the process in step S50 of FIG.

  From the example sentence order rearranging unit 110, an input list including the rearranged example groups (referred to as an input list E) is received. One example group is extracted from the input list E (step S501). It is determined whether or not the extracted example group is an example group with the first rearrangement order (step S502).

  If the rearrangement order is the first example group, the top three sentences are output from the unoutput example sentences of the target example group, and the process proceeds to step S507 (step S503). If the extracted example group is not the first example group, it is determined whether the rearrangement order is the second example group (step S504).

  If the example group has the second order, the upper two sentences are output from the unoutput example sentences of the target example group, and the process proceeds to step S507 (step S505). If the extracted example group is not the second example group, the top one sentence is output from the unoutput example sentences of the target example group (step S506).

  It is determined whether or not example sentences in all example groups have been output (step S507). If there is an example sentence that has not been output, the process returns to step S501. If all example sentences have been output, the process ends.

  Further, there are cases where a plurality of four sets exist in the four examples (for example, the above example sentences (1-i) and (1-ii)). In this case, using the information on the start position of the additional information and the number of characters, it is colored into four sets (when the example sentence is displayed on the screen for each example group, the four sets of characters including the input keyword are highlighted. It is possible to distinguish by performing (display). (Coloring, that is, highlighting, is an example of screen display. The screen display is devised so that the acquired quadruple is conspicuous, such as changing the color of the character itself to a conspicuous color or increasing the size of the character. You can change the color for each part of speech, use different colors for nouns and verbs, use the same colors for nouns and particles, and use verbs. For example, in the example sentence (1) “Display the electronic payment screen on a separate screen”, the additional information “1: Separate-screen + on: Separate screen: noun + particle: 8-4 => 2: Display-display: display: verb-basic-form: 12-4 "creates a quadruple [display, display on screen]. At this time, a combination of the following word, the word description start position and the number of characters in the sentence is acquired from the additional information, the search keyword information, and the part-of-speech information.

[Screen, 9-2] [In, 10-1] [Display, 12-2] [Yes, 12-4]
(Note that the acquisition format in the above example is [word, word description start position−number of characters]. Also, since this example is an example, it is not limited to this format.)
At the time of output, the number of characters is counted from the beginning to the start position, and coloring is performed for each part of speech. This presentation reveals the Japanese grammar (quadruple combinations) that should be noted in the example sentences, and even if there are different quadruples in the same sentence, the difference can be confirmed at a glance. It becomes possible.

  Further, in this example, three, two, and one example sentence are output in order from the upper group of the rearrangement result, but the number of example parts is an example, and is limited to this. It is not a thing.

(Summary)
As described above, according to the example search device according to the present embodiment, the example sentences extracted by the search using the word as a key can be classified without using the case information for each word. Determining case information in advance for each word takes time and cannot process words for which case information is not prepared. However, in the example search device of this embodiment, such a problem may occur. Absent. In addition, the example storage unit 114 of the example search device according to the present embodiment adds a part of speech and dependency information by morphological analysis and syntactic structure analysis of a Japanese document. If processing is performed using existing technology, it can be easily registered in the example storage unit 114. That is, there is no need to add case information to a new word included in the example sentence.

  In addition, the search results are classified by focusing on the four sets of “nouns”, “particles”, “verbs”, and “utilization” that are the basis of Japanese sentences, and example sentences are presented to the user based on the classification (that is, Because a plurality of patterns of example sentences are presented to the user), it is easy to select a suitable example even if the user does not have deep knowledge of Japanese grammar.

  Further, since the example sentence can be searched by word, even a user who does not have the skill to write a grammatically correct Japanese sentence can search for a correct example sentence in Japanese.

  Although one embodiment of the present invention has been described so far, it is needless to say that the present invention is not limited to the above-described embodiment, and may be implemented in various forms within the scope of the technical idea.

  In addition, the scope of the present invention is not limited to the illustrated and described exemplary embodiments, and includes all embodiments that provide the same effects as those intended by the present invention. Further, the scope of the invention is not limited to the combinations of features of the invention defined by the claims, but may be defined by any desired combination of particular features among all the disclosed features. .

DESCRIPTION OF SYMBOLS 1 Example search device 10 Keyword 20 Input device 30 Output device 102 Input / output part 104 Part-of-speech assignment part 106 Example sentence extraction part 108 Example sentence classification part 110 Example sentence rearrangement part 112 Result output part 114 Example storage part 116 Part-of-speech information storage part

Claims (6)

An example storage unit that stores an example sentence, a morphological analysis result of the example sentence, and a structure analysis result of the example sentence;
A part-of-speech giving unit that gives part of speech to the input word;
Example sentences stored in the example storage unit including words whose word-of-speech and word correspondence are the same combination as the input word are used as nouns, particles, verbs including the input word, and patterns of their use. A processing unit to classify based on,
A result output unit for outputting the classified example sentences;
An example search device comprising:   The processing unit further classifies the example groups into which the example sentences are classified into large groups based on part of speech of example sentences to be prioritized that are determined according to the part of speech of the input word. Rearrange the example groups in descending order of the number of examples in the group,
  The example search device according to claim 1, wherein the result output unit extracts and outputs an example sentence for each of the rearranged example groups for each large group.   The processing unit further calculates the number of example groups in each large group, rearranges the large groups in descending order of the calculated number of groups, and sets an example sentence in the example group included in each large group to a sentence length. The example search device according to claim 2, wherein the examples are rearranged in ascending order. For the input word, further comprising a part-of-speech information storage unit storing a correspondence relationship between the morpheme condition and the part-of-speech to be given,
The example part search device according to any one of claims 1 to 3 , wherein the part-of-speech providing unit adds part-of-speech to the input word with reference to the part-of-speech information storage unit. A computer apparatus comprising: an example sentence, a morphological analysis result of the example sentence, and an example storage unit for storing the example sentence structure analysis result in association with each other, a part of speech assignment unit, a processing unit, and a result output unit An example search method to execute,
The part-of-speech giving unit giving a part-of-speech to the input word;
The processing unit includes an example sentence stored in the example storage unit including a word whose part-of-speech and word correspondence is the same combination as the input word, a noun including the input word, a particle, and a verb. A step of classifying based on the pattern of utilization,
The result output unit outputting the classified example sentences;
The example search method characterized by including this. An example storage unit that stores an example sentence, a morphological analysis result of the example sentence, and a structure analysis result of the example sentence in association with each other, a part of speech assignment unit, a processing unit, and a result output unit ,
The part-of-speech giving unit giving a part-of-speech to the input word;
The processing unit includes an example sentence stored in the example storage unit including a word whose part-of-speech and word correspondence is the same combination as the input word, a noun including the input word, a particle, and a verb. A step of classifying based on the pattern of utilization,
The result output unit outputting the classified example sentences;
Example search program to execute

Images