JPH11184890A - Device for preparing dictionary on individual concern - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prepare a dictionary for obtaining individual concern by extracting word and phrase (proper word) indicating an area of his knowledge in which he is interested from text data written by himself. SOLUTION: A document including text data prepared by a specific individual stored in an information storing means 1 is analyzed based on a morpheme so as to be segmented into joints of words and phrases by a text analyzing means 2, and words concurring with a proper noun in a predetermined unit in the text data are extracted as an abstract proper word candidate indicating the knowledge area of the specific individual who prepares the text, data, the proper noun is extracted, and concurrence frequency with each abstract proper word is counted by a analyzing means 3. Then, an abstract proper word extracting means 4 extracts an abstract proper word from abstract proper word candidates by using the concurrence frequency of the proper noun as a reference, and a concrete proper word extracting means 5 extracts a proper noun concurring with the extracted abstract proper word as a concrete proper word indicating concrete concern of the specific individual in the area indicated by the abstract proper word, and a proper word dictionary means 7 holds the extracted abstract proper word and the concrete proper word by making them correspond to each other.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a phrase (hereinafter, referred to as a phrase) representing a knowledge area in which an individual has an interest (including an interest and knowledge).
(Referred to as proper words) from text data written by the individual, and to create a dictionary for grasping the interests of the individual.

[0002]

2. Description of the Related Art In recent years, there has been an increasing need for personal uniqueness data such as individuality and characteristics in order to provide, for example, the selection of attendees to conferences and the selection of mail destinations. recent years,
Even in Japan, the tendency toward individualism has been growing more than before. As a result, the individuality of each individual is being demanded, and the number of places where personalities different from others are actively utilized is increasing.

[0003] On the other hand, conventionally, various types of communication have been greatly limited by physical distance, but with the development of networks and the like, the possibility of communication and collaboration with people who are not always physically close has increased. As a result, for example, it has become necessary to find a person who is most suitable for solving a certain problem, or to find a person who is interested in oneself and become acquainted with him.

[0004] On the other hand, conventionally, for example, when a problem to be solved arises, a method of calling an unspecified majority through a net news or a mailing list or an article posted on the net news Judging from the contents of the website and the home page, a user-led method is adopted in which an e-mail or the like is sent to a person who seems to solve the problem and a request is made for a solution.

[0005] However, in the former case, since the call is made to an unspecified number of opponents, the opponent who is called has little consciousness that the request is made by himself, and whether or not to answer the call depends on the good intentions of the opponent. Who depends heavily and who is truly worthy of the caller. In the case of the latter, since the request is made to a specific individual, it is more likely that the request will be answered than in the former case. Relying on information makes it difficult to find the right person. In this case, it is possible to use a search engine or the like to search and search the homepage, but it is difficult to find a suitable person without actually reading the entire contents, and to find a suitable person However, the burden on the user is very large.

For this reason, in recent years, a system for searching for a person who has an interest or knowledge in a certain area or introducing people who have the same interest has begun to be considered. In such a system, in order for a person to be selected successfully, data on the uniqueness of the knowledge domain, such as what field each individual has knowledge and interest, is required. The term knowledge domain here does not necessarily refer only to purely intellectual things such as academic fields, but also to various things such as those related to hobbies and areas of interest in everyday life. Used.

Japanese Patent Application Laid-Open No. 9-44470 discloses a conventional technique for expressing a region of interest based on a kind of selection, such as introduction of a specialist or a group having the same interest. The present invention seeks to simultaneously express a region of interest and information processing capability in that region. Focusing on the expression of the area of interest in this,
The region of interest of a specific person is represented by a keyword, and the keyword is extracted from the contents of electronic mail transmitted and received by the person. The extraction method is the whole e-mail exchanged (including those not related to the person)
This is based on the ratio of the frequency of appearance of words in an electronic mail transmitted and received by a specific person to the frequency of appearance of words in, and uses a conventional keyword extraction technique for information retrieval. In addition, "Construction of personal network database in distributed personal network utilization support system" (Priority Area Research "Advanced Database" Matsue Workshop Proceedings Vol.1,
pp109-114 (1996)) discloses a technique for extracting words included in a subject of an electronic mail exchanged between two people as a region of interest common to the two people. As described above, conventionally, the region of interest of an individual is simply represented by a single keyword group.

[0008]

However, an experiment was conducted in which individuals were allowed to freely enumerate words representing their area of interest (hereinafter referred to as proper words) as many times as desired.
As a result, the following characteristics were found. (1) For example, when 50 proper words are listed, the person is not interested in the 50 independent regions, and the region of interest is divided into several regions. . (2) Each region of interest is composed of words at different levels of abstraction, from very abstract to very specific. (3) Specific words include many proper nouns such as personal names, place names, novel and movie names, and product names. (4) The words that constitute one region are related to each other, but the relationship is not a strict hierarchical relationship as expressed in a thesaurus, but has a relationship based on various meanings and contexts. It is also often personal, based on the individual's experience and values.

For example, among the proper words given by an individual,
FIG. 8 shows a diagram related to the area of "literature" based on the relationship between the two. That is, even in the area of "literature", there are different areas of interest such as "novel" and "national literature", and the "novel" can be divided into the areas of "sense of novel" and "girl novel". Further, specific objects of interest such as “Junichiro Tanizaki” and “Masashi Yokomizo” have separate or overlapping relations to each region.

[0010] When choosing who is appropriate in discussions and problem solving, specific areas such as an abstract area of "what kind of area are you interested in?" Are needed together. Even if the interests are the same in the abstract domain, the specific objects of interest may be completely different, and the story may not match at all.
Also, for example, even if the same novel is cited as a specific object of interest, for example, one is interested in literature and particularly likes the novel, and the other is not interested in literature, Sometimes you are interested in Africa, and you are interested because the novel describes Africa. In this way, by expressing one's interests in terms of different levels of abstraction, one reinforces each other's meanings. Instead, they must be represented by a set of keywords with different levels of abstraction.

However, having the user enumerate such proper words puts a heavy burden on the user, and it is quite difficult to come up with everything on the spot, and there are many omissions. Also, even if they are enumerated, the result is greatly influenced by the interest at the time of enumeration, and it is quite arbitrary. In addition, putting together related phrases puts a heavy burden on the user. In the above example, only "literature" is shown for simplicity. However, specific interests are extremely complicated because they cover a plurality of areas.

Here, there is a thesaurus as a representation of such word / phrase association, and automatic association or hierarchization using the thesaurus can be considered. However, the thesaurus basically represents the superordinate-subordinate relation of words, but the relation between proper words is not suitable for application because many of them have no hierarchical relation. Furthermore, in the existing thesaurus that is currently available, proper nouns are not described except for some place names, etc., and even if they are listed, words that are increasing every day such as product names and program names It is difficult to be supported.

Under the circumstances described above, there is a need for a method of automatically extracting a set of proper words without burdening the user. Therefore, as shown in FIG. 8, there are actually some hierarchical relationships having several levels, but first, the most specific concrete words at the lowest level, which are particularly important, and Consider the extraction of a two-stage set of more abstract proper words. As an object to be extracted, text information created by the user can be considered as generally showing the interest of a certain person. Here, to reduce leakage, we assume a large amount of text written in various texts, from official ones such as dissertations to everyday and private ones such as e-mail, and for a certain period of time. I do.

Conventionally, as a method for automatically extracting words having a relationship with each other (hereinafter referred to as related words) from text, there is a method based on the co-occurrence frequency of words and phrases. JP-A-6
Japanese Patent Publication No. 168272 discloses that, in order to create a related word for a certain keyword, a document containing the original keyword is compared with the appearance frequency in the entire document database among other words appearing in the document containing the original keyword. There is disclosed an invention in which a keyword that appears frequently in a keyword is related to the original keyword. According to such a method, it is also possible to extract related words other than the upper and lower relations that are not in the thesaurus. However, such a related word extraction method is used in an information search in order to show what related words are used in a large amount of documents including a specified keyword, and to enable a user to narrow down the results. However, even if it is a related word that is effective for narrowing down the results, it does not take into account such points as a set of words with different levels of abstraction.

The present invention has been made in view of the above circumstances, and provides a personal interest dictionary creating apparatus for accurately extracting a set of words representing the uniqueness of an individual's knowledge area and creating the same in a dictionary. It is intended to provide.

[0016]

SUMMARY OF THE INVENTION A personal interest dictionary creating apparatus according to the present invention extracts words representing a specific interest of a specific individual who created the document from a document and summarizes the words in a dictionary. A dictionary creation device, which divides a document including text data created by a specific person stored in an information storage unit into phrasal units by morphological analysis or the like by a text analysis unit, and converts the text into text units by a co-occurrence analysis unit. A word that co-occurs with a proper noun in a predetermined text unit (for example, each document or paragraph) in the data is extracted as an abstract proper word candidate representing a knowledge area of a specific individual who created the text data. The proper noun is extracted and the co-occurrence frequency with each abstract proper word candidate is counted. Then, based on the co-occurrence frequency of the co-occurring proper noun, the abstract proper word extracting means extracts an abstract proper word from the abstract proper word candidate, and further, the specific proper word extracting means A co-occurring proper noun is extracted as a specific proper word indicating a specific interest of the specific individual in an area represented by the abstract proper word, and the extracted abstract proper word and the specific proper word are associated with each other to be unique. Stored in word dictionary means.

Further, in the personal interest dictionary creating apparatus according to the present invention, the co-occurrence analyzing means may identify a word which co-occurs with a proper noun in a predetermined text unit in the text data by a specific individual who has created the text data. And extract the proper nouns, count the number of proper nouns, and use the extracted proper nouns as a basis to extract abstract proper word candidates from the abstract proper word candidates. A specific noun is extracted, and a specific noun which co-occurs with the extracted abstract proper word is extracted by the specific proper word extracting means. The extracted abstract proper word and the specific proper word are associated with each other and stored in the proper word dictionary means.

The set of proper words in the proper word dictionary created in this way can be regarded as related words for the individual, and therefore can be used not only in personal selection and the like but also in information retrieval and the like. It can be used as related word information to support a search tailored to the individual.
For example, the use of related words in information search may be used to expand a query or may be used to narrow down results. In the case of query expansion, more results are obtained by expanding and searching the query with related words. On the other hand, in the case of narrowing down the results, narrowing down to only those that match the user's search intention depending on what words are used as related words of the query.

The proper words extracted in the present invention can be used for both of these purposes. In the former case, specific words such as proper nouns have high specificity of meaning, so by searching for documents that contain proper nouns that are related words of the query, the words of the query are included. You can search for necessary documents even if they are not. In the latter case, only documents intended by the user can be searched according to what proper nouns co-occur with documents satisfying the query. Furthermore, the proper words extracted by the present invention can be used in any other system that supports the intellectual activity of the individual, which requires information on the region of interest of the individual.

[0020]

An embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows the configuration of a personal interest dictionary creation device according to the present embodiment. The personal interest dictionary creation device includes an information storage unit 1, a text analysis unit 2, a co-occurrence analysis unit 3, an abstract unique It includes a word extraction unit 4, a specific proper word extraction unit 5, a proper word registration unit 6, and a proper word dictionary 7. The function units 2 to 6 and the function units 21, 22, 31, and 32, which will be described later, in the present embodiment are configured by executing a program installed in advance using computer hardware resources.

The information storage unit 1 has a device such as a magnetic disk device for storing information in a readable and writable manner, and stores a plurality of documents including text data written by a specific individual. Note that the information storage unit 1 may be installed at a location on the network away from the text analysis unit 2 or the like, or may be configured by a plurality of devices. Further, the above-described document may include data such as a picture or a figure in addition to the text data.

The text analysis unit 2 analyzes text data in a document stored in the information storage unit 1 and cuts out independent words (nouns, adjectives, verbs, etc.) in the text data. In the present embodiment, the text analysis unit 2
Is provided with a morphological analysis unit 21 and a word processing unit 22, and performs a process of extracting a noun from text data. The morphological analysis unit 21 divides the text data in the document stored in the information storage unit 1 into words by performing morphological analysis, and performs a process of adding part of speech information to each word. Note that morphological analysis is widely known as a basic technology of natural language processing. For example, “Basic technology of natural language processing” (Hirogo Nomura, The Institute of Electronics, Information and Communication Engineers, 1988)
And information processing (Vol. 30, No. 10, 1989), “3.1 Morphology” and the like.

The word processing unit 22 processes unknown words and compound words based on the analysis result of the morphological analysis unit 21. For example,
Since there are many foreign words and new words in proper nouns, unknown words are regarded as proper nouns, and part of speech information of proper nouns is given. In addition, since the continuation of kanji nouns and the continuation of katakana nouns are considered to be compound words, the part of speech information is given as one noun collectively. In the present embodiment, the above-described configuration is based on the premise that an existing morphological analysis system is used.
The present invention is not limited to this, and the morphological analysis unit 21 and the word processing unit 2
2 may be performed simultaneously. At this time, a short phrase surrounded by, for example, "" or "" may be used as a proper noun based on a predetermined rule.

In addition, a proper noun dictionary is provided in the text analysis unit 2 in addition to the morphological analysis dictionary, and unknown words and compound words are matched with proper nouns in this dictionary, and those that match are proper nouns. You may do so. In the present embodiment, the morphological analysis is used as the analysis content of the text analysis unit 2. However, the present invention is not limited to this, and a deeper analysis such as syntax analysis is performed in accordance with information required by the co-occurrence analysis unit 3. Analysis may be performed.

The co-occurrence analysis unit 3 provides, for each text unit,
A word co-occurring with a proper noun is extracted as an abstract proper word candidate representing the knowledge area of the creator of the document data, and the co-occurrence frequency of the co-occurring proper noun is counted for each abstract proper word candidate. Here, in the present embodiment, “co-occur”
Means "coexist in the same text unit". However, the present invention is not limited to this, and may refer to, for example, a case having a grammatical dependency relationship. In this case, the text analysis unit 2 may perform a deeper analysis such as a syntax analysis.

The text unit is one document in the present embodiment, and is regarded as “co-occurring” if both are present in the same document. The present invention is not limited to this,
Any unit may be used as long as it is a predetermined text unit. For example, one sentence, one paragraph, or the like may be used as a text unit. When the document stored in the information storage unit 1 is a structured document, the document may be a range surrounded by a specific tag. In addition to the structural features, a certain number of rows,
The number of characters may be used. When "co-occurrence" occurs when there is a grammatical dependency relationship, it is preferable to use one sentence as a text unit.

In this embodiment, a noun is considered as a word co-occurring with a proper noun. In this embodiment, the stem of the sa variant is also regarded as a sa noun, and is included in the noun. Is excluded. Hereinafter, when an abstract proper word candidate in the present embodiment is written as a noun without notice, it refers to a noun in the above range. In addition, the present invention is not limited to this, and the point is that any word that can be an individual proper word may be used, for example, broadly independent words, all nouns and verbs, etc. It may be a noun only, or an adjective or an adjective depending on the purpose.

In the present embodiment, the co-occurrence analysis unit 3 includes a word extraction unit 31 and a co-occurrence counting unit 32. The word extracting unit 31 extracts the proper noun and the noun in the document for each document,
Each list is extracted without duplication, and a proper noun list and an abstract proper word candidate list for each document are created. Here, eliminating duplicate means that, for example, in the creation of a proper noun list, if the same proper noun appears twice or more in the same document, the second and subsequent times are ignored. Co-occurrence counting section 3
2 counts the co-occurrence frequency of co-occurring proper nouns (the number of co-occurring documents in the present embodiment) for each abstract proper word candidate from the proper noun list and the abstract proper word candidate list of all documents, Create a co-occurrence list.

The abstract proper word extraction unit 4 selects a word representing a knowledge region unique to the creator of the document from the abstract proper word candidates in the co-occurrence list based on the co-occurrence frequency of each co-occurring proper noun. I do. The criterion for this selection is simply that the number of co-occurring proper nouns is more than one, and the total number of appearing documents is more than how many. In addition, an evaluation formula including these numerical values is created, and the evaluation value is equal to or greater than a predetermined value.
Alternatively, a criterion such as a percentage from a larger evaluation value to a percentage may be used. Further, factors other than the number of co-occurring proper nouns and the number of documents of each proper noun may be further added as described later. Any method based on numbers may be used.

The specific proper word extracting unit 5 selects a specific proper word related to each abstract proper word extracted by the abstract proper word extracting unit 4. Here, as a selection criterion, for example, among the proper nouns in the co-occurrence list, those whose co-occurrence frequency is greater than a predetermined number are selected. In addition, the selection criterion may be, for example, selecting n from the largest number of co-occurring documents, or selecting n% of the total, or a combination other than the co-occurrence frequency.

The proper word registration unit 6 performs a process of registering the extracted abstract proper word and concrete proper word as a set in the proper word dictionary 7 of the document creator. Here, the frequency of each proper noun, the evaluation value calculated by the abstract proper word extraction unit 4 and the specific proper word extraction unit 5, and the like may be registered together. The proper word dictionary 7 is configured by a device such as a magnetic disk device that stores information in a readable and writable manner, and stores a set of the above-described abstract proper words and specific proper words as a dictionary for each individual.

Next, a process of creating a unique word dictionary for a specific individual, which is performed by the personal interest dictionary creating device, will be described with reference to FIGS. The information storage unit 1
Is assumed to store N documents created by Mr. A. First, the morphological analysis unit 21 of the text analysis unit 2 morphologically analyzes text data in all documents stored in the information storage unit 1 (step S).
1). That is, the morphological analysis unit 21 first reads text data from a certain document D1 one sentence at a time, divides the sentence into words, and adds part of speech information to each word. For example, if there is a sentence in the text “Chinatown in Penang has a considerable traditional landscape.” As shown in FIG. 3, the sentence is divided into words by morphological analysis. The part of speech information is added. When the analysis of all the sentences of the document D1 is completed as described above, the morphological analysis unit 21 moves to the next document D2 and performs the same analysis.

When the morphological analysis has been completed for all the documents N, the word processing unit 22 processes unknown words and compound words (step S2). For example, when the analysis result shown in FIG. 3 is obtained, when the processing of the word processing unit 22 is performed, as shown in FIG. 4, the unknown word “Penang” is a proper noun and a series of nouns. "China" and "town" are put together into a noun "Chinatown". The processing of unknown words and compound words is performed in order from the beginning of the document D1 to the document DN.

Next, in the word extraction unit 31 of the co-occurrence analysis unit 3, a proper noun list and an abstract proper word candidate list are created (step S3). The word extracting unit 31 first extracts all proper nouns and nouns from the first document D1 without duplication, and creates a proper noun list PNL1 and an abstract proper word candidate list AIL1. For example, "Chinatown in Penang has a lot of traditional scenery. Georgetown's store architecture is called a shophouse.
It is a form that serves both as a residence and a store. ), The proper noun list contains the proper nouns "Penang" and "Georgetown", and the abstract proper word candidate list contains "" Chinatown "".
The words "tradition", "landscape", "store architecture", "shophouse", "house", "shop", and "form" are included. When the list is created for the document D1, the word extraction unit 3
1 moves to the next document D2 and similarly creates a list, and creates each list for all documents N.

Next, the co-occurrence counting unit 32 counts co-occurring proper nouns for each abstract proper word candidate, and creates a co-occurrence list COL (step S4). The operation of the co-occurrence counting unit 32 will be described in detail with reference to FIG.
(Steps S11 and S12), and the document D1
It is determined whether or not there is a word in the small and medium proper word candidate list (step S13).
If there is a word, the document D1
(W) from the beginning of the abstract proper word candidate list AIL1
Take it out (step S14).

Next, the extracted word W is the co-occurrence list CO
L (step S15), and if not, the word is added to the list (step S16). If there is, the word is confirmed in the proper noun list PNL1 of the document D1 (step S17). If there is no word, the process proceeds to the next abstract proper word, while if there is a word, one word (PN) is extracted from the head of the proper noun list PNL1 of the document D1 (step S18). Then, it is checked whether or not this word (PN) is in the co-occurrence proper noun of the term W in the co-occurrence list (step S19). If not, the PN is added to the list and its frequency is set to 1 On the other hand (step S20), if there is, the frequency is incremented by +1 (step S21).

When the list of all proper nouns in the proper noun list is changed by the above-described processing, the process moves to the next word in the abstract proper noun candidate list, and similarly, from the first word in the proper noun list, Add to the starting list or add frequency. Further, when this processing is completed for all the nouns in the abstract proper word candidate list of the document D1, the process proceeds to the next document D2, and the abstract proper word candidate list AIL of the document D2 is executed.
2 and the proper noun list PNL2. Thereafter, similarly, the abstract proper word candidate list and the proper noun list up to the document N are processed, and the co-occurrence list COL is updated. For example, when the abstract proper word candidate list and the proper noun list of the document D1 and the document D2 are as shown in FIG. 6A, the co-occurrence list COL at the time when the processing on the document D1 and the document D2 is completed is completed. FIG. 6B shows a part thereof. The number shown in FIG. 6B is the frequency.

When the co-occurrence list COL for all the documents N is created as described above (step S4), the abstract proper word extraction unit 4 uses the number of appearing documents of each proper noun in the co-occurrence list. A word representing a knowledge region unique to the creator of the document is selected from among the abstract proper word candidates (step S
5). When the extraction of the abstract proper words is completed, the specific proper word extracting unit 5 extracts the specific proper words related to each selected abstract proper word (step S6). Here, among the proper nouns in the co-occurrence list, those whose co-occurrence frequency is larger than a predetermined number are selected. For example, FIG.
When a co-occurrence list COL partially shown in FIG. 7 is created, for example, with a threshold value of 40, a set of an abstract proper word and a specific proper word as shown in FIG. It is extracted by the operation of the unit 4 and the specific proper word extracting unit 5.

When the selection of the specific proper word is completed in this way, in the proper word registration unit 6, the extracted abstract proper word and the specific proper word are set as a set and registered in the proper word dictionary 7 of the creator of the document. (Step S7), a series of dictionary creation operations ends. The proper word dictionary 7 created in this way is
For example, use as effective information that indicates the area of interest of a person in a sort of person selection system that selects people who solve problems based on individual interests and knowledge, and introduces people with similar interests Can be. In the present embodiment, the personal interest dictionary creating apparatus of the present invention has been described as creating a dictionary that permanently stores and retains information. However, this dictionary is used as a buffer that temporarily retains information. Alternatively, for example, the result of extracting the proper word may not be registered as permanent information but may be used directly in the application.

In the above embodiment, the co-occurrence counting unit 3
2, a co-occurrence list is created by counting the co-occurrence frequency of proper nouns co-occurring from the proper noun list and the abstract proper word candidate list, and the abstract proper word extraction unit 4 calculates the co-occurrence frequency based on the co-occurrence frequency of each proper noun. Thus, a word representing a knowledge area unique to the creator of the document is selected from the abstract proper word candidates in the co-occurrence list. However, as another embodiment of the present invention, the co-occurrence counting unit 3 is used.
2 generates a co-occurrence list by counting the number of co-occurring proper nouns for each abstract proper word candidate from the proper noun list and the abstract proper word candidate list of all the documents. Alternatively, a word representing a knowledge area unique to the creator of the document may be selected from abstract natural word candidates in the co-occurrence list based on the number of co-occurring proper nouns. That is, an abstract proper word to be selected for registration may be selected according to the number of proper nouns related to the abstract proper word.

[0041]

As described above, according to the present invention,
Individual candidate nouns are selected from among abstract proper word candidates that co-occur with proper nouns as specific proper words that express specific interests of a specific individual based on the number of the proper nouns and the co-occurrence frequency. A set of words representing the uniqueness of a region can be accurately extracted and created in a dictionary. For this reason,
Using this dictionary, for example, it is possible to appropriately select a person and an e-mail address, and to smoothly execute tasks related to personal intellectual activities.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a personal interest dictionary creation device according to an embodiment of the present invention.

FIG. 2 is a flowchart illustrating a processing operation according to an embodiment of the present invention.

FIG. 3 is a diagram illustrating an operation of a text analysis unit according to an embodiment of the present invention.

FIG. 4 is a diagram illustrating an operation of a text analysis unit according to an embodiment of the present invention.

FIG. 5 is a flowchart illustrating an operation of a co-occurrence counting unit according to an embodiment of the present invention.

FIG. 6 is a diagram illustrating an operation of a co-occurrence counting unit according to an embodiment of the present invention.

FIG. 7 is a diagram illustrating operations of an abstract proper word extraction unit and a specific proper word extraction unit according to an embodiment of the present invention.

FIG. 8 is a diagram illustrating a conventional problem. 1 ... information storage unit 2 ... text analysis unit 3
... Co-occurrence analysis unit, 4 ... Abstract proper word extraction unit, 5.
..Specific proper word extraction unit, 5 ... proper word registration unit, 7
..Property dictionary, 21 ... morphological analyzer, 22 ..
・ Word processing unit, 31 ・ ・ ・ Word extraction unit, 32 ・ ・ ・ Co-occurrence counting unit,

Claims (3)

[Claims] 1. An apparatus for creating a personal interest dictionary that extracts, from a document, words representing specific interests of a specific individual who has created the document and compiles the words into a dictionary. An information storage unit that stores a document including the text, a text analysis unit that analyzes text data in the document, and a text that co-occurs with a proper noun in a predetermined text unit in the text data. It is extracted as an abstract proper word candidate representing the knowledge area of a specific individual,
Further, a co-occurrence analysis means for extracting the proper noun and counting the co-occurrence frequency with each abstract proper word, and extracting an abstract proper word from the abstract proper word candidate based on the co-occurrence frequency of the co-occurring proper noun. An abstract proper word extracting means for extracting, and a specific noun which co-occurs with the extracted abstract proper word as a specific proper word representing a specific interest of the specific individual in an area represented by the abstract proper word A personal interest dictionary creating apparatus, comprising: proper word extracting means; and proper word dictionary means for holding the extracted abstract proper words and concrete proper words in association with each other. 2. A personal interests dictionary creation device for extracting words indicating a specific interest of a specific individual who created the document from a document and compiling the extracted words into a dictionary, wherein text data created by the specific individual is An information storage unit that stores a document including the text, a text analysis unit that analyzes text data in the document, and a text that co-occurs with a proper noun in a predetermined text unit in the text data. It is extracted as an abstract proper word candidate representing the knowledge area of a specific individual,
A co-occurrence analyzing unit that extracts the proper noun and counts the number; an abstract proper word extracting unit that extracts an abstract proper word from the abstract proper word candidate based on the extracted proper noun number; Specific proper word extracting means for extracting a proper noun co-occurring with the abstract proper word as a specific proper word indicating a specific interest of the specific individual in a region represented by the abstract proper word; A personal word dictionary creating device, comprising: a proper word dictionary means for storing a proper word and a specific proper word in association with each other. 3. The personal interest dictionary creating apparatus according to claim 1, wherein the information storage unit stores a plurality of documents created by the same individual, The co-occurrence analysis unit, the abstract proper word extraction unit, and the specific proper word extraction unit perform the processing on a plurality of documents related to the same individual. A personal interest dictionary creation device characterized in that an abstract proper word and a specific proper word relating to the same individual extracted from the same are held in association with each other.