JPH06187329A - Natural language processor - Google Patents

Abstract

PURPOSE:To simplify generating and maintaining operation for a dictionary while preventing the dictionary from increasing in size and to automatically extract information necessary for machine processing from meaning sentences described for people. CONSTITUTION:The processor is equipped with a dictionary 10 having a common information storage part 11 for storing common terms between information for the people and the information for machine processing, a retrieving device 22 which retrieves the dictionary 10 on the basis of retrieval information, and an analytic device 23 which analyzes the meaning, etc., of meaning sentences obtained by the dictionary retrieval by analyzing information on the reading and parts of speech, morpheme, and syntax, and the information necessary for the machine processing is extracted from a text described for the people on the basis of the common items of the dictionary 10.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a natural language processing device used in a word processor or the like, and more particularly to a natural language processing device for performing a search or the like using a dictionary used in a document processing device such as a Japanese word processor.

[0002]

2. Description of the Related Art Since Japanese has many character types, a large number of input methods have been put into practical use as compared with the input methods of European languages. Among the typical input methods used in Japanese word processors, Kana-Kanji conversion input including Romaji-Kana-Kanji conversion input is most often used, and the conversion process also includes bunsetsu conversion that incorporates Japanese grammar analysis, and The continuous phrase conversion, which is an extension of this, is the mainstream.

At present, there are roughly two types of dictionaries used in document processing devices such as Japanese word processors.

Human dictionary: An electronic version of a booklet dictionary that is normally used in daily life. The content is roughly the same as the booklet dictionary, including pictures and drawings. is there.

Machine dictionary: A dictionary for Japanese (natural language) processing such as machine translation and kana-kanji conversion. Usually, information is stored in a format that is not easily understood by humans. There is.

Hereinafter, the reason why the machine dictionary has not been used as a human dictionary in natural language processing (in other words, a machine dictionary has been developed separately from the human dictionary) will be described.

Problems Concerning Contents of Human Dictionaries Up to now, dictionaries of one or several dictionary editors (Lexi
Most of them were created by cographer over a long period of time. As a result, problems such as omission of item description, item bias, typographical errors, ambiguity, and contradiction cannot be avoided. Further, some information necessary for machine processing of a language cannot be directly obtained from a general human dictionary.

Problems with Retrieval of Human Dictionaries Human dictionaries are so useful that a person who has grammatical information and other basic knowledge about the language can use them by repeating the training of "drawing a dictionary". Is. In fact, it is said that there is no dictionary that foreigners who learn the language for the first time can use it satisfactorily. Therefore, from the viewpoint of information retrieval, the structure of the human dictionary is completely unsuitable for processing by a machine without prior knowledge of the language.

Problems Concerning Contents of Machine Dictionary From the viewpoint of the machine dictionary, the natural language processing technology in the early days did not perform technically high-level processing,
I didn't need the definitions or examples of words in the human dictionary. Alternatively, even if it is recognized that it is necessary, it is not possible to use the information up to that point due to hardware limitations such as storage capacity and search speed.

Therefore, at the present stage, the human dictionary is not actively used in natural language, and at most, editing work of translation results in machine translation and homophones (homophone synonyms / homophones) in kana-kanji conversion are performed. In the selection process, etc.), the meaning of words, translations, synonyms, antonyms, etc. are presented to human beings and they are only used in a very passive manner.

[0011]

Therefore, in the natural language processing apparatus used for such a conventional Japanese word processor, a human dictionary and a machine dictionary are constructed and used separately. Therefore, some items must be created separately, even though there are items common to both dictionaries, which makes the dictionary bloat, inefficient creation / maintenance work, and dictionary reconstruction work. There were problems such as the difficulty of.

The problem will be specifically described below.

Problems of Enlargement of Dictionary Capacity Both human and machine dictionaries are becoming enormous in size. For example, it is said that there are 100,000 examples of the AI dictionary for a word processor-dedicated machine, and 300,000 to 500,000 cases of the large one. However, there are many cases in which the same phoneme cannot be selected correctly, which makes it practical. Not reached the level. Therefore, it is clear that a larger amount of examples is needed to improve the accuracy of the conversion.

Problem of dictionary management Some large-scale databases have been put into practical use until now, but since updating the data takes more money and time than expected, maintenance is stopped and eventually it will not be used. There are many things that have been lost. Therefore, managing two large-scale dictionaries, one for humans and one for machines, is too time and financially burdensome.

Problem of multiple registration of common items Originally, notation, reading, grammar, meaning, example, accent, etc.
There are many common items in the human dictionary and machine dictionary.
However, under the present circumstances, even if the contents are exactly the same, they must be registered in all dictionaries. Therefore, if more advanced natural language processing and a more sophisticated dictionary that can be easily used by foreigners are realized, it is possible that the number of common items will increase.

Problem of Loss of Information For example, taking the example dictionary for AI conversion as an example, the example dictionary for AI conversion is information in which necessary information for internal processing is extracted and classified from examples. If the viewpoint is slightly different, it is almost useless for another processing. In this way, there is not a little information included in the examples themselves but lost at the stage of creating the machine dictionary, and as a result, use the current machine dictionary as it is as a dictionary for more advanced language processing. Is impossible, and a huge amount of work of rebuilding the dictionary is required.

In addition to the above-mentioned problems, a dictionary for natural language processing is based on the idea of "preparing all the information necessary for processing in advance". It is a well-known fact that it is practically impossible. The same is true for human dictionaries, for example, it is not uncommon to use English original books together, and the meaning of words may still be unknown.

In this case, generally, a method of supplementing necessary information by using a user-defined dictionary or a rudimentary learning function is used, but there is a limit in a range and a degree which can be dealt with, and it is essential. Problem has not been solved yet.

Therefore, in the next-generation natural language processing technology,
Something that has been digitized but not a dictionary, such as
Document files, computerized newspapers / magazines, online databases such as BBS, etc. are used as sources of information, and the necessary information can be used by accessing these sources in the same way as the current dictionaries. There must be.

An object of the present invention is to prevent the dictionary from becoming too large by unifying the common parts of the two types of dictionaries that currently exist, a human dictionary and a machine dictionary, including word meaning sentences. It is to be able to improve the efficiency of the work of creating and maintaining and the work of rebuilding the dictionary.

[0021]

The means of the present invention are as follows.

The dictionary stores information common to human search information and machine processing information as common items. The structure of the dictionary is essentially the meaning of sentences written for humans. The information necessary for machine processing can be automatically extracted from (text). The search means searches the information stored in the dictionary based on the search information. For example, the dictionary is searched based on keywords to search for necessary information, and if the search fails, there is no corresponding item. Is a search device that returns the information.

The analyzing means extracts the necessary information from the dictionary by the searching means, and also extracts the information necessary for the machine processing from the text written for humans based on the common items of the dictionary. For example, it is an analysis device that extracts information such as reading and part-of-speech information and information necessary for each internal processing from a word meaning sentence by a form / syntactic analysis processing. Further, a judgment means for judging whether or not the necessary information is obtained by the output of the analysis means may be provided.

[0024]

The operation of the means of the present invention is as follows.

First, it is assumed that the dictionary stores retrievably common items of information for human search and information for machine processing.

When the user inputs the search information in this state, the information stored in the dictionary is searched by the search means based on the input search information, the necessary information is extracted from the dictionary, and the dictionary is further searched. Information necessary for machine processing is extracted from texts written for humans based on common items.

Therefore, by realizing a dictionary in which common items are unified, it is possible to prevent the size of the dictionary from being enlarged and to simplify the work of creating and maintaining the dictionary. Information necessary for machine processing can be automatically extracted from the word meaning sentence described in.

[0028]

EXAMPLES Examples will be described below with reference to FIGS.

Description of Principle First, the basic concept of the present invention will be described.

FIG. 1 is a diagram showing an actual description example of a dictionary. In FIG. 1, these description examples do not describe the definitions of words, that is, meanings and concepts, but merely rephrasing the same with different expressions. There are many such descriptions in human dictionaries, and it is not always possible to obtain information such as definitions of meanings and concepts necessary for machine processing and usage of words. Therefore, human dictionaries are said to be unsuitable for machine processing.

However, for example, when the expression "king's place" is not useful as information, "king's place" is analyzed using a natural language processing technique such as morphological analysis / syntactic analysis. If it can be correctly recognized that the expression is composed of "king" (noun), "no" (case particle), and "rank" (noun), then "king" and "rank" are used as search keywords. By continuing the search, you can get more useful information. Similarly, in the case of "drinking sake", the expression "drinking sake" is "drinking sake (Western sake / beer)" if you continue searching for "drinking sake".
It turns out that it contains the expression and meaning.

This process is basically the same as the "dictionary lookup" method, especially when learning foreign languages such as English and French. By the way, if you look up a dictionary to look up the meaning of a word or expression and the meaning is not clear, it is common to look up the dictionary again using the words that make up the definition as keywords. The meaning is clarified in such a way.

When the word meanings of the dictionary are used as the information source, an expression pattern peculiar to the dictionary as shown in FIG. 2 (for example, a kind of ~, a Japanese word of ~, a Chinese expression, ~, etc.) is collected. By creating rules, it is possible to reduce the burden of advanced natural language processing based on parsing.

Therefore, in the present invention, in the dictionary search, by creating a combined dictionary in which common items of information for human beings and information for machine processing are unified, the enlargement of the size of the dictionary is prevented, and Realizes simplification of dictionary creation / maintenance work.

As a characteristic of the common dictionary, not only the structure of the dictionary, but also the information necessary for machine processing is automatically extracted from the meaning sentences such as definitions and examples originally described for humans. To do so. In addition, while applying the technology for analyzing the word meaning sentence described above and using only the dedicated dictionary in which the conventional natural language processing describes the necessary information in advance,
A text existing on a terminal or a network (this text refers to any kind of electronic information) is extracted, and information necessary for human or machine processing is extracted from the text. That is, away from the restriction of the conventional dictionary of "defining all necessary information in advance", the user prepares the minimum necessary dictionary on the terminal, and if necessary, the super-large scale on the network. Realize an online dictionary that accesses a database to obtain information.

Embodiment FIG. 3 to FIG. 5 are views showing an embodiment of the natural language processing apparatus.

First, the structure will be described. FIG. 3 is a diagram showing the configuration of the dictionary 10 of the natural language processing apparatus created based on the above basic principle.

In FIG. 3, the dictionary 10 is necessary for a common information storage unit 11 that stores common items of information for humans and information for machine processing, and for machine processing such as machine translation and kana-kanji conversion. However, it is composed of a machine-dedicated information storage unit 12 that stores machine-dedicated information that is not necessary for humans, and a human-dedicated information storage unit 13 that stores information for helping human understanding.

In the common information storage unit 11, reading / notation /
In addition to including information such as part-of-speech, it also includes word meaning sentences in a dictionary such as definitions and examples. It is not necessary to process the word meaning sentence for machine processing, and the information may be stored in a format that is easily understood by humans. This is because when used in machine processing as described in the principle description, necessary information is extracted using a natural language processing technique. In this case, in order to improve the search speed and reduce the capacity of the dictionary, it is necessary to appropriately hold the information. The machine-dedicated information storage unit 12 describes information that is necessary for machine processing, such as machine translation and kana-kanji conversion, but not for humans. Further, in order to improve the processing speed, basic information frequently used in the mechanical processing device may be stored in this portion. The human-dedicated information storage unit 13 stores information such as drawings and pictures that cannot be used by machine processing but help human understanding.

Note that the dictionary 10 in FIG. 3 is merely a conceptual diagram of the structure of the dictionary, and when actually implementing the dictionary, even if each information is stored in a physically continuous storage area, it is stored separately. You may do it.

FIG. 4 is a block diagram of the dictionary search device 20 for searching the dictionary 10. In the figure, reference numeral 21 denotes a dictionary search control device 21 including a CPU for controlling the entire dictionary search and performing various data processing including a text search process described later. The dictionary search control device 21 is stored in a ROM (not shown). Various operations of the natural language processor are controlled according to the micro program. The dictionary search control device 21 receives search information (keywords, etc.) from the dictionary search control device 21, and searches the dictionary 10 based on the received search information, and the word meaning sentence obtained by the dictionary search. On the other hand, information such as reading (notation) / part of speech and an analysis device 23 that analyzes meanings by morpheme / syntactic analysis processing, a storage device 23 that stores the analyzed information in a storage area, and a storage area of the storage device 23. The evaluation device 24 is configured to evaluate whether or not necessary information is obtained by referring to the stored information. A search control device 30 that controls the entire search is connected to the dictionary search control device 21, and the search control device 30 is composed of, for example, a terminal such as a personal computer.
To request the dictionary search to control the whole and receive the dictionary search information from the dictionary search device 20. A display control device 31 that displays the retrieved information on a display or the like is connected to the search control device 30.

The search device 22 receives the search information (keyword, etc.) from the dictionary search control device, searches the dictionary 10 based on the information, and if the information can be detected, the information etc. (dictionary address) necessary for subsequent analysis. Etc.) to the dictionary search control device 21. The analysis device 23 extracts information necessary for each internal processing from the word meaning sentence by information such as reading (notation) / part of speech and form / syntactic analysis processing. In this case, the information extracted from the word meaning sentence includes, for example, in the case of machine translation, synonyms (synonyms) and antonyms for changing meaning (concept), translation, or expression (style). The evaluation device 24 refers to the storage area of the storage device 23 to determine whether or not necessary information is obtained, and when it is determined that the necessary information is insufficient, the evaluation device 24 selects from the information stored in the storage area. , Extract search information (keywords etc.) when searching again,
The control is returned to the dictionary search control device 21 so as to search again based on the information.

Next, the operation of this embodiment will be described.

Operation of dictionary search device 20 When the search is started, the search device 22 causes the dictionary search control device 2 to operate.
Receives keywords etc. from 1, and dictionary 1 based on that information
Search for 0. If the information can be retrieved, the information such as the dictionary address necessary for the subsequent analysis is returned to the dictionary retrieval control device 21, and if the retrieval fails, the information that there is no corresponding item is returned. If the search fails, the dictionary search control device 21
Returns the information that there is no corresponding item to the search device 30 that controls the entire search, and ends the process.

When obtaining information necessary for humans, the dictionary search is terminated when the information search is successful, and the display control device 31 displays the detected information on a display or the like. In this case, when the user gives an instruction to search for more detailed information or related information, the same processing as the search for machine processing described below is continued.

In order to obtain information necessary for machine processing, when the dictionary search is successful, the analysis device 23
Information necessary for each internal processing is extracted from the word meaning sentence by reading (notation), part-of-speech information, and morpheme / syntactic analysis processing. In this case, the information extracted from the word meaning sentence includes, for example, in the case of machine translation, synonyms (synonyms) and antonyms for changing meaning (concept), translation, or expression (style).

The extracted information is stored in the output information storage area of the storage device 25, and control is returned to the dictionary search control device 21.

The evaluation device 24 also refers to the storage area of the storage device 25 to determine whether the necessary information has been obtained. If it is determined that it is insufficient, search information such as keywords for re-searching is extracted from the information stored in the storage area, and the dictionary search is performed again based on that information. The control is returned to the control device 21. However, if it is determined that valid information cannot be obtained even if the search is continued for a further time, the search process is interrupted and the control is returned to the dictionary search control device 21. If it can be determined that the information necessary for the processing has been sufficiently obtained, the search end information is returned to the dictionary search control device 21, and the processing ends.

For example, in the dictionary search device 20 shown in FIG. 4, when searching for the information "drink", "drink"
(Noun), "wo" (case particle), and "drink" (verb) are just paraphrases of "drinking", and when you want to get more detailed information, you can use the searched information "drink" Need to analyze. Therefore, the analysis device 23 analyzes the morpheme of the information “drinking sake” and first connects the noun “sake” and the case particle “o” to the verb “drinking” clause. 10 is used for analysis, and the noun “sake” and the verb “drink” are searched using the dictionary 10 to increase information. Here, the dictionary 10 is necessary for machine processing such as common information storage 11 having common information about reading, notation, part of speech, etc. as shown in FIG. 3 and machine translation such as grammatical information and kana-kanji conversion. A machine-dedicated information storage unit 12 that stores machine-dedicated information and a human-dedicated information storage unit 1 that stores information such as diagrams and pictures to help humans understand
3 and are accessed by the search device 22 and the analysis device 23. That is, when searching, first, a common item in the common information storage unit 11 is searched, and as a result of the search,
When the common information alone does not provide the necessary and sufficient information, access to the dictionary 10 described below is performed. If the necessary information cannot be obtained from the common information alone, for example, when a person wants to know information about a certain word, he / she can access the place where the word of that word is stored in the human-only information section 13 to obtain the information. If there is no information in the machine translation or the like, the machine-dedicated information section 12 is accessed to retrieve the information related to the machine translation. When the search is not completed by the dictionary 10, for example, the database on the network may be accessed by the network connected to the dictionary search device 20 that uses the dictionary 10.

Returning to FIG. 4, the dictionary 10 is searched to increase the information as described above, and the result is evaluated by the evaluation device 24. When the result is not suitable for the purpose of the search, the analysis device 23 analyzes again. Try again to accumulate more detailed information. When this result is evaluated again by the evaluation device 24 and a search suitable for the purpose is performed, the search result is output to the user side and the search operation is terminated.

Operation of Search Control Device 30 FIG. 5 is a flowchart of a text search process in the search control device 30, and this flow is the dictionary search device 20.
Is a process for searching the text on the network when sufficient information is not obtained. Here, the text means all information sources that have already been digitized.

First, in step S1, the dictionary search control device 2
The dictionary search information from 1 is received, and it is determined in step S2 whether or not the search is successful. When the dictionary search is successful, it is judged that it is not necessary to perform the text search on the terminal and the network, and the processing of this flow is ended. When the dictionary search is not successful, the text search on the terminal and the network is performed. Therefore, in step S3, it is determined whether or not there is text in the terminal. If there is text in the terminal, a search is executed using the text as an information source in step S4, and if there is no text in the terminal, the process proceeds to step S6 to perform a text search on the network. Next, in step S5, it is determined whether or not the text search using the text of the terminal as an information source is successful, and when the search is successful, it is determined that it is not necessary to perform the text search on the network, and the process of this flow is performed. Finish. Also, step S
If the text search is not successful in step 5, it is determined in step S6 whether or not there is text on the network in order to perform text search on the network. The search is executed using the text as the information source, and the processing of this flow ends. If there is no text on the network in step S6, it is determined that all text searches have been completed, and the processing of this flow ends.

By executing the above flow, the search control device 30 specifically performs the following operation.

It is assumed that the dictionary search control device 21 returns to the search control device 30 information that the search has failed or that the extracted information is insufficient. In this case, the search control device 30 first checks whether or not there is text on the terminal, and if there is, executes the search using the text as an information source. The processing method of the search is the same as that of the dictionary search described above.

If the search is successful, the process is terminated, and if the search is unsuccessful, the search process for the text on the network is executed. However, when there are a plurality of texts on the terminal, the processing may be changed so as to search them in order, and this also applies to the text search on the network.

By the way, in reality, the retrieval process differs depending on the type of text. For example, in the case of text on a terminal, the search method differs depending on whether the target is an ascii format file or an original file format such as a dedicated word processor. Similarly, in the text search on the network, it is necessary to convert the search information into an SQL language or a general-purpose database query (Query) language at present, but in this case as well, a search processing function according to the search target is equipped. do it.

As described above, the natural language processing apparatus of this embodiment controls the entire dictionary search and the dictionary 10 that unifies common items of information for humans and information for machine processing, and will be described later. A control device 21 that performs various data processes including a text search process, a search device 22 that searches the dictionary 10 based on the search information, and information and morphemes such as reading and part-of-speech for a word meaning sentence obtained by the dictionary search. The dictionary is provided with the analysis device 23 that analyzes the meaning and the like by the structure analysis process, and the information necessary for machine processing is extracted from the text described for humans based on the common items of the dictionary. While preventing the size from bloating,
It is possible to simplify the creation and maintenance of dictionaries,
In addition, information necessary for machine processing can be automatically extracted from a word meaning sentence (definition / example) written for human beings. Furthermore, while conventional natural language processing uses only a dedicated dictionary that describes necessary information in advance, it searches for text (all kinds of digitized information) existing on a terminal or network. However, it can extract the necessary information (for human or machine processing) from it.

Particularly, in the natural language processing apparatus of this embodiment,
From the viewpoint of creating and maintaining the dictionary, the following effects can be obtained by unifying the common parts of the existing two types of dictionaries, the human dictionary and the machine dictionary, including the word meanings of the dictionaries.

Prevention of enlargement of dictionary: Since common items can be unified, even if the dictionary becomes extremely large in future, at least the capacity can be made much smaller than that of the current method. .

Efficiency of dictionary creation / maintenance work ...
By unifying the dictionaries, the work of creating and maintaining dictionaries
Significantly reduce.

Elimination of restructuring work of the dictionary ........ If this natural language processing device is used, even if a higher level natural language processing than the current one is required, the dictionary for realizing the processing is re-created. Don't be bothered by the huge amount of work to build from. Therefore, it is possible to easily introduce a more advanced processing technique.

Further, from the viewpoint of the natural language processing technique, since a dictionary for both human and machine is realized,
There are the following advantages.

Utilization of a variety of information sources ...... By using this natural language processing apparatus, it is not necessary to prepare a machine dictionary for information necessary for natural language processing in advance, Can be utilized.

Customization of dictionary: As described above, the dictionary used by each user needs only the minimum necessary information for maintaining the processing efficiency. This overcomes the shortcomings of the current schemes, in which the dictionary capacity grows inexhaustibly.

Realization of online dictionary ・ ・ ・ ・ ・ Above,
Is characterized by, for example, an individual user having a minimal dictionary on the terminal and accessing a large database on a networked host computer if more detailed or specialized information is needed. This is a basic technology to facilitate the realization of an online dictionary.
The natural language processing device having the above characteristics can be applied to, for example, the grammar correction support function of a document processing device such as a word processor. Here, as the current grammar correction support function,
The following functions have been put to practical use. In other words, spell check function: Checks whether the spelling is correct (including the check of homophone selection for kana-kanji conversion).

Grammar check function: Matches the subject and predicate, checks how to use the particle "teniha", etc.

Stylistic check function ・ ・ ・ ・ ・ Unified "do ・ masu" at the end of the sentence, or a function to check honorific expressions.

Semantic check function by inference ・ ・ ・ ・ ・ For example, if there is a sentence "Penguin flew in the sky," a knowledge base (common sense) "Penguin is a bird but cannot fly" By using, we detect that this sentence is inconsistent and cannot actually occur.

In addition to the above, almost no grammar proofreading support function for checking whether or not the content of a sentence is correct is realized, but if this natural language processing apparatus is applied, the fact of the content shown in the sentence The function of checking whether the relationship is correct can be realized by using a large-scale database on the network. For example, suppose now that the sentence "XX company has announced a new type of personal information device technology Pascal in New York" is correct. In this case, an information database regarding product announcements of XX company is available on the network (for example, BBS's Nikkei Mix apple.japan forum).
If the news subcommittee etc.) exists, if the database is searched and the search results are analyzed by the analysis device proposed by the present invention, the place where the XX company announced is Chicago.
It would theoretically be possible to correct the product name as "Newton".

In this embodiment, an example of creating a dictionary having a structure as shown in FIG. 3 is shown, but common items of information for human search and information for machine processing are stored in a searchable manner. Any dictionary can be used as long as it is a dictionary that does.
For example, the information source is not limited to the meaning of sentences such as a national language dictionary, not to mention dictionaries such as medical dictionaries, six law encyclopedias, encyclopedias, electronic newspapers and magazines, and, for example,
It is possible to use a specialized book in a specific field such as "Inside Macintosh" as it is as an information source without forming a special database. However, when documents are to be searched, the method is limited in the method of regularizing the expression pattern as in the case of word meaning sentences in the dictionary (that is, the pattern matching method used in the first-generation machine translation). Obviously, there is a need for more advanced natural language processing techniques (mainly parsing). Although no particular parsing technique is specified here, LFG (Lexical), which has been put into practical use in recent years, is not specified.
Functional Grammar) and GPSG (Generalized Phrase)
Applied technology such as Structure Grammar) or applied technology such as Extended Context Free Grammar. Further, in the present embodiment, the natural language processing device can be applied to a document processing device such as a word processor, but as long as it is a device having a function of searching a dictionary, all other devices such as a personal computer and a work device. It goes without saying that it can also be applied to stations.

[0071]

According to the present invention, a dictionary is provided which stores information common to human information and machine processing information as a common item in a searchable manner, and is based on the common item of the dictionary. By extracting the information necessary for machine processing from the text written for human beings, it is possible to prevent the size of the dictionary from expanding and create / create a dictionary.
It is possible to realize simplification of maintenance work, and it is possible to automatically extract information necessary for machine processing from a word meaning sentence written for human beings.

[Brief description of drawings]

FIG. 1 is a diagram showing a description example of a dictionary of a natural language processing device.

FIG. 2 is a diagram showing a representation example of word meaning sentences in a dictionary of a natural language processing device.

FIG. 3 is a diagram showing a configuration example of a dictionary of a natural language processing device.

FIG. 4 is a block configuration diagram of a dictionary search device of a natural language processing device.

FIG. 5 is a flowchart of a text search process of the natural language processing device.

[Explanation of symbols]

 10 dictionary 11 common information storage unit 12 machine dedicated information storage unit 13 human dedicated information storage unit 21 dictionary search control device 22 search device 23 analysis device 24 evaluation device

Claims (1)

[Claims] 1. A dictionary that stores information common to human information and machine processing information as a common item, and a search that searches information stored in the dictionary based on the search information. Means for extracting necessary information from the dictionary by the searching means, and analyzing information necessary for machine processing from text described for humans based on common items of the dictionary. A natural language processing device characterized in that