JPH0782497B2 - Document processor - Google Patents

Description

TECHNICAL FIELD The present invention relates to a document processing apparatus for automatically performing format division and classification of a document created by a user or a document input from an electronic mail by content.

<Prior Art> In a document processing device such as a Japanese word processor,
When creating a relatively standard document such as a business document,
It is often the case that you can use a document created in the past as it is or create a new document by just partially modifying it.
This is also a great advantage of word processors. Therefore, the user classifies the documents created by him / her according to a certain classification system, adds a classification index (document name) that symbolizes the document contents to this, and stores it in the memory. It is used to retrieve a desired document from the memory and call it.

<Problems to be Solved by the Invention> However, the document classification method in the above-mentioned conventional word processor is based on manual work based on the judgment of the user. Therefore, in order to perform accurate classification, the user needs to display the classification system list. There is a problem that it is necessary to fully understand it, and different users have different criteria for judging items not listed. Therefore, it is very difficult to classify the created documents uniformly and efficiently according to their contents. Further, when the number of documents classified into the same category increases, the number of similar document names increases, and the document contents cannot be clearly determined only by the document name, which makes it impossible to perform an accurate search. For example, a document with the document name "Celebratory letter" can be broadly classified into (1) Opening, Opening congratulatory letter, (2) Newly constructed congratulatory letter (3) Eiko, promotion letter, etc. The contents of each sentence are greatly different from each other, and even in the same classification, there is a difference in the expressions of the preceding sentence, the main sentence, and the end sentence that compose the sentence, for example, when opening a store and when opening a store.

Therefore, an object of the present invention is to cause the document processing apparatus itself to automatically classify a document in multiple ways based on the cause of the generation, the purpose of creation, the format paragraph pattern, etc., and add and register an appropriate classification index. It is an object of the present invention to provide a novel document processing device capable of surely and efficiently searching for a desired document.

<Means for Solving the Problems> In order to achieve the above object, the document processing apparatus of the present invention is a format control means for dividing an arbitrary document input by a user into a plurality of sentences by identifying a format paragraph pattern. And a keyword extraction means for analyzing a word contained in the sentence to extract a keyword candidate that symbolizes the meaning of the word, and a topic paired with a keyword and a topic name representing the cause of the sentence containing the keyword. A topic / keyword relationship table storage means for storing a keyword relationship table, and for each keyword, a predetermined keyword and another related keyword appearing in a sentence including the predetermined keyword are related to an important topic name related to the predetermined keyword. Rule table storage means for storing the rule table, and whether the keyword candidates extracted by the keyword extraction means From the above, select the keyword and its topic name in the topic / keyword relation table, and the topic name is important depending on whether there is a related keyword based on the above rule table in the sentence preceding the word corresponding to the selected keyword. In addition to determining whether it is a topic name, if not, topic classification / analysis means that changes to the important topic name according to the above rule table, conventional expressions and keywords appearing in the sentence, and the sentence Document purpose / keyword, idiomatic expression relation table means for storing the object and the object of interest, and idiomatic expressions contained in the sentence are extracted using a dictionary, and the document object is extracted by the extracted idiomatic expression and the keyword candidate. -Document purpose classifying means for selecting candidate creation purposes of documents by referring to keywords and idiomatic expression relation tables, and each topic name The consistency between the topic / document purpose correspondence table storage means showing the suitability of each combination of document purposes, the topic determined by the topic classification / analysis means, and the document purpose selected by the document purpose classification means is described above. The input document is divided and classified into each of the divided sentences by providing a document concept control means for inspecting based on the topic / document objective correspondence table and determining the combination of the topic / document objective forming the document concept. While adding a topic name and a document purpose name as an index and storing them in a storage device, a document can be created by retrieving a sentence from the storage device based on a given topic name or the like.

<Operation> The document input by the user is first divided into a plurality of sentences by the format control unit by identifying the format paragraph pattern, and the words in this sentence are analyzed by the keyword extraction unit to determine the meaning content. The symbolic keyword candidates are extracted. In the topic / keyword relationship table storage means,
A topic / keyword relationship table in which a keyword and a topic name indicating a cause of a sentence including the keyword are paired is stored, and the rule table storage means stores a predetermined keyword for each keyword and a sentence including the predetermined keyword. A rule table for associating other related keywords that appear with important topic names related to the predetermined keyword is stored.

Therefore, the topic classification / analysis means selects a keyword and its topic name in the topic / keyword relationship table from the keyword candidates extracted by the keyword extraction means, and a sentence preceding the word corresponding to the selected keyword. If there is a related keyword based on the above rule table, it is determined whether the topic name is an important topic name.If not, change to the important topic name according to the rule table. Do. On the other hand, the document purpose classifying means extracts the idiomatic expressions contained in the sentence, and based on the extracted idiomatic expressions and the keyword candidates found by the topic classifying / analyzing means, the idiomatic expressions appearing in the sentence. A document creation purpose candidate is selected by referring to the document purpose / keyword and idiomatic expression relation table means that store the keyword and the purpose of the sentence in pairs.

Finally, the document concept control means determines the consistency between the topic obtained by the topic classification / analysis means and the document purpose selected by the document purpose classification means, whether or not the combination of each topic name and each document purpose is appropriate. Is checked based on the topic / document purpose correspondence table that represents, and the combination of the topic / document purpose that constitutes the document concept is determined.

The topic name and document purpose name thus determined are added as classification indexes to each of the divided sentences,
While being stored in the storage device, a text corresponding to a given topic name or the like is read from the storage device at the time of inspection, whereby a new document can be easily created.

<Examples> Hereinafter, the present invention will be described in detail with reference to illustrated examples.

FIG. 1 is a schematic diagram of the configuration of a Japanese word processor which is an example of the document processing apparatus of the present invention. The central processing unit 1 is provided with various means and tables described below, and the document data input from the input device 2 is pseudonym. Edited by converting kanji,
While automatically classifying according to the document content, the input document data and the processed document data are displayed on the display device 3 and stored in the auxiliary storage device 4.

The central processing unit 1 divides the input document data into a preamble, a main sentence, a final sentence, etc. by a format paragraph pattern such as a paragraph structure, indentation, and an inscription, and a keyword dictionary from the divided main sentences. 6, a plurality of words are selected, keyword extracting means 7 for extracting a superordinate concept word of the selected words, that is, a keyword candidate, and an important topic for expressing a topic which is a cause of the document from the extracted keyword candidates Topic analysis means 11 for finding a keyword and analyzing the sentence in front of the word corresponding to the found keyword to check the validity of the topic 11
Equipped with.

Further, the central processing unit 1 has the conventional expression extracting means 13
Is referred to by the document purpose classification means 15 for determining a candidate for creating a document based on the idiomatic expression extracted from the sentence by referring to the idiomatic expression dictionary 12 and the extracted keyword candidates, and the topic analysis means 11 From the selected topics, a proper one is finally selected by referring to the topic hierarchy table 16 (see FIG. 4), and the selected topic and the creation purpose candidate determined by the document purpose classification means 15 are matched. Sex
The document purpose correspondence table 17 (see FIG. 5) is inspected, and a document concept control means 18 for finally determining an appropriate combination of the two is provided.

The finding of an important keyword in the topic analysis unit 11 is determined by the topic classification unit 9 first determining a topic suitable for the keyword candidate extracted by the keyword extraction unit 7 based on the topic / keyword relationship table 8. Rule table from combinations of keywords and topics
This is done by choosing the most important keywords listed in 10. Further, the document purpose classifying means 15 refers to the document purpose / keyword / conventional expression relation table 14 based on the conventional expressions extracted by the conventional expression extracting means 13 and the keyword candidates extracted by the keyword extracting means 7. Determine the candidate for document creation purpose. Then, the central processing unit 1 adds each topic name, document purpose name and format classification name finally determined by the document concept control means 18 as an index for each text paragraph of the main sentence data divided by the format control means 5. Then, it is stored in the auxiliary storage device 4 and registered. On the other hand, the central processing unit 1 searches the auxiliary storage device 4 for a document corresponding to the index input from the input device 2 at the time of search, and displays a document as a search result on the display device 3.

The automatic classification processing of documents by the Japanese word processor having the above configuration will be described below.

When the user inputs a document from the input device 2, the central processing unit 1 causes the format control unit 5 to input the input document to the second document.
As shown in the figure, the sentence is divided into the preceding sentence, the main sentence, the end sentence, and the like, and the following analysis is performed to determine the document concept for the divided main sentence. Now, it is assumed that the divided main sentence is the sentence 20 shown in FIG. The vertical line in the figure indicates reference by a keyword dictionary or a relationship table. Then, the keyword extracting means 7 uses the keyword dictionary 6 to convert each word in the sentence 20 into a keyword. In this example, seven keywords “company, organization, change, store, post, inauguration, Kei” are extracted as shown in the keyword 21 column in the figure. Subsequently, the topic classifying means 9 selects “change” and “appointment” as topic candidates 22 from the extracted seven keywords using the topic / keyword relationship table 8 and selects keywords for the topic candidates “change”. For "reform", the keywords "position" and "appointment" are extracted from the topic candidate "appointment", respectively.

Next, the topic analysis means 11 analyzes the preamble sentences of the words corresponding to the keywords for the keywords described in the rule table 10 among the extracted combinations of the topic candidates and the keywords, and determines the validity of the topic candidates. Inspect sex. In the rule table 10 described above, for example, the following rule is registered for the keyword “reform”.

This rule states that if there is a word converted to the keyword "organization" in front of some words in the sentence converted to the keyword "transformation", the topic "change" is not simply a change of address or phone number. It means that it should be interpreted as a foreign topic “personnel change”. Also, for the topic "appointment", rules are registered such that the word (department manager) converted to the keyword "post" is required in front of the word converted to the keyword "appointment". Carefully select topic candidates based on this.

Next, the conventional expression extraction means 13 extracts the conventional expression that is included in the sentence and symbolizes the purpose of the sentence by using the conventional expression dictionary 12. In the example of FIG. 3, as shown by the underline in the sentence 20, "accepting place" and "had been trained" are extracted as idiomatic expressions peculiar to the letter sentence. Then, the document purpose classifying means 15 uses the document purpose / keyword and the idiomatic expression relation table 14, and both of the above two idiomatic expressions are expressions found in the document purpose “celebration letter” or “sympathy letter”, Since the keyword "Kei" indicates a document purpose "celebration letter", the first candidate for the document purpose is determined to be a "celebration letter" as shown in the column of the document purpose candidate 23 in FIG.

Further, the document concept control means 18 refers to the topic hierarchy relation table 16 shown in FIG. 4 from the above-mentioned carefully selected topic candidates, and determines a more specific one as a final topic candidate. In the case of the example sentence, the topic “initiation” is more specific (subordinate concept) than the topic “personnel change”, and thus the topic is determined to be “invocation”. Next, referring to the topic / document purpose correspondence table 17 shown in FIG. 5, the validity of the combination of the topic candidate and the document purpose candidate is checked. In the case of the example sentence, the combination of the topic "Inauguration" and the document purpose "Celebration letter" is valid, but in the case of the combination indicated by the cross in Fig. 5, it is considered invalid and the topic / document purpose candidate is reexamined. Done.

Thus, the topic name and document purpose name are, for example, "appointment",
When the "Celebration letter" is finally determined, the central processing unit 1 adds the topic name and the document purpose name as a classification index to each paragraph of the sentence data of the main sentence divided by the format control means 5, and stores the auxiliary storage. It is stored in the device 4 and registered. On the other hand, when creating a new document using the document registered in this way, the user inputs a desired topic name or document purpose name from the input device 2 as a search condition. Then, the central processing unit 1 reads out the document data having the classification index corresponding to the search condition from the auxiliary storage device 4 and displays it on the display device 3. Therefore, the user can modify the displayed document or refer to it to easily and efficiently create a new document.

In the above embodiment, the various processing means 7, 11, 15, 18 provided in the central processing unit 1 are provided with various dictionaries 6, which make the processing efficient.
Since 12, the tables 8, 9, 14, 16, 17 and the auxiliary processing means 9, 13 are provided, the automatic classification of documents can be further speeded up.

<Effects of the Invention> As is apparent from the above description, the document processing apparatus of the present invention includes a format control unit that divides a document input by the user into a plurality of sentences and a keyword candidate from the words in the sentence. The keyword and its topic name in the topic / keyword relationship table (a table that stores a keyword and a topic name indicating the cause of the sentence containing this keyword as a pair) from the keyword candidates that are extracted The rule table in the sentence in front of the word corresponding to the selected keyword (for each keyword, the predetermined keyword and other related keywords appearing in the sentence including this predetermined keyword as important topic names related to the predetermined keyword). Depending on whether there is a related keyword based on the table stored independently), the topic name above Is a key topic name, and if it is not, a topic classification / analysis means that changes to a key topic name according to the above rule table, and a convention extracted from a sentence using a dictionary Document purpose / keywords and idiomatic expression relation table based on dynamic expressions and the above keyword candidates, and document purpose classification means for selecting candidate creation purposes for documents, topics determined by the above topic classification / analysis means, and the above document purposes. The consistency with the document purpose selected by the classification means is checked based on the topic / document purpose correspondence table,
A document concept control means for deciding a combination of topics and document purposes that form a document concept is provided, and the determined topic and document purpose names are added to the divided sentences as a classification index and stored, while being given. Since the document corresponding to the classification index can be searched to create a document, the created document can be quickly and multi-directionally automatically registered and registered, and then the desired document can be searched accurately and efficiently. , With this as a reference, new documents can be created easily and efficiently, which greatly contributes to dramatic improvement in document management.

[Brief description of drawings]

FIG. 1 is a schematic diagram showing the configuration of a Japanese word processor which is an embodiment of the present invention, FIG. 2 is a diagram showing an example of format division according to the above embodiment, and FIG. 3 is a concept showing a concrete example of classification processing according to the above embodiment. Fig. 4 is a diagram showing an example of hierarchical relation of topics,
FIG. 5 is a diagram showing an example of availability of a combination of topics and document purposes. 1 ... Central processing unit, 2 ... Input device, 4 ... Auxiliary storage device,
5 ... Format control means, 7 ... Keyword extraction means, 11 ... Topic analysis means, 15 ... Document purpose classification means, 18 ... Document concept control means.

Claims (1)

[Claims] 1. Format control means for identifying a format paragraph pattern of an arbitrary document input by a user and dividing the document into a plurality of sentences, and analyzing a word included in the sentence to symbolize the meaning of the word. A keyword extraction means for extracting keyword candidates, a topic / keyword relationship table storage means for storing a topic / keyword relationship table that pairs a keyword and a topic name indicating the cause of a sentence containing the keyword, and for each keyword , A rule table storing means for storing a rule table for associating a predetermined keyword and other related keywords appearing in a sentence including the predetermined keyword with an important topic name related to the predetermined keyword, and the rule extracting means extracted by the keyword extracting means. From the keyword candidates, select the keywords in the topic / keyword relationship table Pick out the pick name, and determine whether or not the topic name is an important topic name, depending on whether there is a related keyword based on the rule table in the sentence in front of the word corresponding to the selected keyword,
If it is determined to be no, the topic classification / analysis means for changing important topic names according to the above rule table and the purpose of the document that stores the idiomatic expressions and keywords appearing in the text as a pair with the purpose of the text keyword,
The conventional expression relation table means and the conventional expression contained in the sentence are extracted using a dictionary, and the document purpose / keyword and conventional expression relationship table are referred to by the extracted conventional expression and the above keyword candidates, Document purpose classification means for selecting candidate creation objectives, topic / document purpose correspondence table storage means for indicating the suitability of each topic name and each document purpose combination, and the topic obtained by the topic classification / analysis means and the document purpose The consistency with the document purpose selected by the classification means is checked based on the above topic / document purpose correspondence table,
A document concept control means for determining a combination of topics and document objectives forming a document concept is provided, and the input document is divided, and the topic name and the document objective name are added to each of the divided sentences as a classification index. A document processing apparatus which is capable of storing a document in a storage device and retrieving a sentence from the storage device based on a given topic name or the like to create a document.