JPH08314980A - Constructing method, displaying method and displaying device for document data base - Google Patents

Abstract

PURPOSE: To realize a subject extraction and expression which is capable of properly grasping document contents without constructing a word dictionary and knowledge concerning grammar and without limiting the field of the contents of the document. CONSTITUTION: An independent word group in a cooccurrence relation capable of suggesting the subject is extracted from a given document to calculate the number of the cooccurrence times of each independent word and the cooccurrence word to it to express the strength of cooccurrence in a network by cooccurrence probability obtained by this. As a result of this, the means of a document is extracted without constructing a grammar dictionary of a large scale and by making up for grammatical dictionary necessary for preparing a summary by referring to this network output by a person, subject extraction and summary generation from this network output are additionally enabled. Thus, a sentence can be grasped more minutely than information of only a key word.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document database for extracting specific words from newspaper articles, technical documents, literary writings, etc., and relating the words to each other in a network format.

[0002]

2. Description of the Related Art Recently, an increasing number of newspaper articles, technical documents, and literary writing documents have been digitized.
Many of these computerized documents do not have keywords and abstracts created by the author, and various attempts have been made to automatically create abstracts by extracting the subject on an electronic computer.

As a method of extracting a word or a sentence indicating a subject from a specific document, there are firstly a method of utilizing the appearance frequency of a keyword and secondly, a method of linguistic analysis.

The former is based on the theory that "in general, what is desired to be emphasized in a document is repeatedly stated, and frequently occurring words are likely to indicate the subject of the document." However, in the independent words, "and",
"That", "this", "early", "(such a thing)",
There are words that appear as linguistic functions (connectives, adnominals, adverbs, formal nouns, prefixes, suffixes, part of demonstrative pronouns, etc.) such as "ka (yowai)" and "(tsuyaya) poi". . Therefore, it is not possible to determine the important word only by the appearance frequency. It is necessary to exclude such language function words and extract a word that has a meaning independently and represents a feature of a document. For that purpose, first, an important word list having a large number of leaves is prepared in advance, and then the appearance frequency of a single word in a document is calculated to determine a keyword. In determining this keyword, the result of linguistic analysis described below may be used. Then, a summary is generated based on the words that appear frequently in this keyword group.

However, since this method cannot deal with a document in which words are paraphrased by synonyms, it is difficult to accurately extract the means simply by using the frequency of appearance of the keywords.

Further, in order to deal with the synonyms, it is necessary to prepare a large-scale thesaurus (synonym dictionary), which causes a problem that the system load increases.
On the other hand, the latter linguistic analysis is a method of deciding important words by paying attention to the structure of connectives and sentences and setting unique parameters for these, and analyzing the relation between sentences and the logical structure of documents.

According to this method, a noun phrase accompanied by "ha" or "ga" as an adjunct word or a word at the end of a sentence ending with "da" or "is" is an important word. And For example, in the case of a sentence "weather will be fine tomorrow", the noun phrase "weather" accompanied by "ha" is an important word.

This method is effective if the content of the target document is limited in the field (for example, limited to only technical documents), but like a literary writing document, it is possible to use "one day", "long time", etc. Different specific parameters are needed when judging the development of a story from the conjunction and the description of the main character's action.

[0009]

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and (1) does not build knowledge about a word dictionary or grammar, and (2) limits the field of document contents. Without (3) human beings, it is an object to automatically perform subject extraction and expression generation that enable accurate grasp of document contents without human intervention such as document preprocessing.

[0010]

The present invention first executes an independent word extracting step for extracting an independent word from a given document. In this step, an independent word is extracted from the sentence. The "independent word" in the present invention refers to a word that has a meaning by itself. For example, in the case of a sentence "I want to be a teacher of cooking", "cooking", "teacher" and "naru" are independent words.

Next, a co-occurrence table creating step for creating a co-occurrence table for recording the co-occurrence words and the co-occurrence frequency of each of the extracted independent words is executed. The term "co-occurrence" as used in the present invention refers to a relationship between two independent words that exist in the same sentence at the same time. That is, when multiple independent words are present in the same sentence at the same time, those independent words are said to be co-occurring with each other or have a co-occurrence relationship. Further, when there are two independent words that have a co-occurrence relationship with each other, one independent word is called a co-occurrence word for the other independent word.

Specific examples of the co-occurrence table are shown in FIGS. 1 and 2. FIG. 1 shows an example in which a co-occurrence table is created based on two sentences, sentence 1 “I want to be a cook's teacher” and sentence 2 “I am a cooking school teacher”. In the figure, “cooking”, “teacher”, and “nar” are extracted as independent words of sentence 1, and “cooking” as independent words of sentence 2,
"School" and "teacher" are extracted. The co-occurrence words on the right side of the figure having a co-occurrence relationship with these independent words are extracted and the co-occurrence frequency is calculated. For example, the independent word “cooking” exists in both sentence 1 and sentence 2, and the co-occurrence word “teacher” is once in sentence 1 and in sentence 2 for this independent word “cooking”. It has appeared twice and has been co-occurred twice in total. Therefore, the co-occurrence number of the co-occurrence word “teacher” for the independent word “cooking” is 2.

FIG. 2 shows an example of creating a co-occurrence table based on the sentence "I go to a cooking school to become a cooking teacher." Next, a co-occurrence probability calculating step for calculating a co-occurrence probability representing the strength of the co-occurrence relationship is executed based on the co-occurrence table. Here, the co-occurrence probability is
When there is an independent word _{w i} appeared, is the probability the co-occurrence word _{c j} with respect to _{W i} is to co-occur with _{w i P (c j / w} i), can be calculated by the following equation (1).

[0014] _{P (c j / w i)} = M (c j / wi) / N wi ····· (1) However, _{w i:} independent words _{(1 ≦ i ≦ N B,} N B: one document Total number of words in the content) c _j : co-occurrence word for independent word w _i (1 ≦ J ≦ N
_{b (wi), N b (} wi: the number of co-occurrence word for the _wi)
M (c _j / w _i ): the number of co-occurrences of the independent word w _i and the co-occurrence word c _{j corresponding} thereto N _wi : the number of appearances of the independent word w _i The co-occurrence probability calculated by the above equation (1) is Independent word w _i
Is a conditional probability of occurrence of w _i and shows a semantic connection from w _i to the co-occurrence word c _j therefor, so this information is a one-way link such as “independent word w _i → co-occurrence word c _j ”. Can be expressed by

Next, the present invention executes a co-occurrence count expected value calculation step for calculating an expected value of the co-occurrence count. The expected value E (c _j / w _i ) of the number of times of co-occurrence can be calculated by the following equation (2).

E (c _j / w _i ) = N _s · {1- (1-P _cj ) ^m − (1-P _wi ) ^m + (1-P _cj ) m · (1-P _wi ) ^m } (2) However, N _s : total number of sentences in one document N _i : total number of independent words m: average number of independent words in one sentence P _cj = N _cj / N _I P _cj : of co-occurrence word c _j the occurrence probability _{P wi = N wi / N I} P wi: independent words _{w i} occurrence probability _m-N I / N _s then, the expected value _{E (c} j / _{w i)} and the actual co-occurrence number of times M (c
_j / w _i) is compared, and a subject suggestive independent word group selection step of extracting a combination of an independent word satisfying a certain condition and a co-occurrence word corresponding thereto is executed. As a certain condition, the following equation (3) can be exemplified.

M (c _j / w _i )> E (c _j / w _i ) (3) Then, if the above equation (3) is satisfied, the expected value E (c _j / w _i )
_The actual co-occurrence number M (c _j / w _i ) is larger than _i ), and it can be said that there is a semantic connection between the independent word w _i and the co-occurrence word c _j . However, the case is extremely small word equivalent number N _B with respect to the total independence number of words N _I in the document, for example, when repeated often with a small word amount, such as the story of a child-friendly, semantic connection is (3) above even if very weak
Many combinations of independent words and co-occurrence words that satisfy the condition of the expression will appear.

Therefore, considering the average number of appearances of one word, it may be determined that the co-occurrence relation is strong only when the following expression (4) is satisfied. M (c _j / w _i )> E (c _j / w _i ) + α · N _I / N _B (4) α can be experimentally obtained, for example.

FIG. 3 shows the total number of independent words is 5984 words and the word quantity is 11 words.
FIG. 9 is a graph showing the relationship between the number of co-occurrences M (c _j / w _i ) set to α = 1.5 using 98-word co-occurrence data and its expected value E (c _j / w _i ). . Fig. 4 shows α
= 2.5, FIG. 5 shows a case where α = 0.5 under the same conditions. The points in the plots in FIGS. 3 to 5 indicate combinations of co-occurring independent words, and the points above the solid line in each figure satisfy the above equation (4). That is, if α is increased, only information having a stronger co-occurrence relationship can be extracted, and if it is reduced, information having a weaker co-occurrence relationship can be extracted. The setting of this α may be set to 1.5 as an initial value, and may be arbitrarily changed by the user.

Next, the present invention executes a co-occurrence network generation step to set a link from an independent word to a co-occurrence word for it and generate a co-occurrence network. That is, the semantic connection of the independent word groups having a strong co-occurrence relation is expressed by a link from the independent word w _i to the co-occurrence word c _{j corresponding} thereto, and the above-mentioned co-occurrence probability P (c _j / w _i ).
Is reflected in the display form of the link to indicate the strength of the co-occurrence relationship, and a co-occurrence network visually representing the subject of the document is generated and displayed. This co-occurrence network visually detects, for example, a character string representing each independent word in the independent word group selected in the subject suggestive independent word group selection step and a character string representing each independent word in these co-occurrence relationships. The co-occurrence line that connects the two can be formed on a screen such as a CRT.

The co-occurrence line is, for example, the line type depending on the magnitude of the co-occurrence probability of a certain independent word and the co-occurrence word corresponding thereto
A line or arrow whose color, shade, direction, length, or thickness is determined can be exemplified. Further, the magnitude of the co-occurrence probability may be visually distinguishable by a three-dimensional expression.

An example of the generated co-occurrence network is shown in FIG. Such co-occurrence network is CR
It can be displayed and visualized on a T screen or an output device such as a printer, so that the subject of the document can be easily grasped visually.

Here, an arrow and a line are used as the display form of the link, and the thickness of the arrow and the line is used as the co-occurrence probability P (c _j
/ W _i ) is divided into the following three levels depending on the strength.

[0024]

[Table 1] Such a table may be held in the storage device of the electronic computer in the form of a table (for example, a level table).

The circled numbers in FIG. 6 correspond to the levels in Table 1 above. In the figure, it can be seen that the independent word “Mama” has the strongest link to the co-occurrence word “I”.

As described above, according to the present invention, the subject of a document can be extracted without constructing a large-scale grammar dictionary. Furthermore, because the semantic connection of independent words that can suggest the subject in a network form can be visually expressed, it is possible to perform a manual search and an automatic search by a computer very efficiently.

In addition, a person can refer to the network output to supplement the grammatical knowledge necessary for creating the abstract, and the subject can be further extracted and the abstract can be generated from the network output. It is possible to understand detailed sentences.

Further, in the present invention, since the constant (α in the equation (4)) is made variable in extracting the subject suggestion word, the number of links to be extracted (displayed) can be freely increased or decreased according to the user's request. . Therefore, it is possible to create a database and create a co-occurrence network regardless of the content of a document, from a story such as an old tale composed of a small amount of words to a technical document having many words.

Further, according to the present invention, a user can refer to the co-occurrence network to identify a necessary original sentence and easily perform an operation for acquiring the original sentence. The following method using the apparatus is provided.

That is, the present invention searches for one or more co-occurrence networks generated in the co-occurrence network generation step, and includes one or more co-occurrence networks containing arbitrary keywords as independent words. A document search step of extracting and displaying on the screen, and reading the original text of a document corresponding to an arbitrary co-occurrence network among the one or more co-occurrence networks displayed on the screen in the document search step and displaying the screen. Perform the text reference step to be displayed.

In the document search step, a search formula input unit for inputting a search formula for specifying a keyword and a search execution command input unit for inputting a search execution command are displayed on the screen of the display device. You may At this time, the user uses the input device such as a keyboard to input the search formula and inputs the search execution command to the search execution command input unit. When the search expression and the search execution instruction are input, the document search step searches one or more co-occurrence networks generated in the co-occurrence network generation step, and searches for the keyword specified by the search expression. The containing co-occurrence network is extracted. The extracted co-occurrence network is displayed on the screen by a CRT or the like.

When the co-occurrence network is displayed on the screen,
For example, the independent word that matches the keyword may be emphasized or displayed separately from others. As a method of emphasizing or distinguishing, a method of displaying an independent word that matches a keyword with a color, typeface, size, or the like different from other independent words can be exemplified.

When the original sentence is displayed on the screen in the original sentence reference step, there are a method of displaying the entire original sentence on the screen and a method of displaying a part of the original sentence on the screen. When displaying a part of the original sentence on the screen, display the sentence or character string containing the keyword input in the document search step, or the sentence or character string containing the keyword and the sentence or character string before and after it. May be. At this time, a sentence including only the keyword or the word may be emphasized and displayed. Examples of the method of emphasizing include a method of displaying only a keyword or a sentence including the keyword in a shaded manner, a method of displaying in reverse, a method of making a character color, a typeface, and a size different from other characters. .

When a new keyword (second keyword) is input in the original text reference step in addition to the keyword (first keyword) input in the document search step, the second keyword is input. Display the sentence that contains. As a method of specifying the second keyword,
A method of designating the display coordinates of the second keyword by the position designating means can be exemplified.

When requesting the reference of the original text,
A method of designating the display coordinates of the co-occurrence network corresponding to the original sentence by the location designating means can be exemplified. Furthermore, since there is a possibility that there are multiple sentences containing keywords in the original sentence, one sentence of the multiple sentences is displayed on the screen, and the keyword (search formula) and the sentence containing the keyword are included in the original sentence. The number that indicates how many there are, the number that indicates which sentence is the currently displayed sentence in the sentence that includes the keyword, and the number that is before the currently displayed sentence that includes the keyword. A pre-sentence display command input unit for inputting a display command and a next sentence display command input unit for inputting a display command of a sentence next to the currently displayed sentence among the sentences including the keyword may be displayed on the screen. .

At this time, the user uses a numerical value indicating how many sentences including the keyword are in the original sentence, and a numerical value indicating the number of the sentence currently displayed among the sentences including the keyword. By referring to, it is possible to recognize that the sentence including the keyword exists before or after the currently displayed sentence. Then, when the user inputs a preamble display command to the preamble display command input unit, the sentence before the currently displayed sentence among the sentences including the keyword is displayed on the screen and the sentence including the keyword is currently displayed. The number that indicates which sentence the current sentence is is 1
Is decremented. If the user inputs a next sentence display command to the next sentence display command input unit, a sentence next to the currently displayed sentence among the sentences including the keyword is displayed on the screen and the current sentence among the sentences including the keyword is displayed. The numerical value indicating the order of the displayed sentence is incremented by one.

Further, in the co-occurrence network generation step, when the generated co-occurrence network is displayed on the screen, a constant change instruction input unit for inputting an instruction to change a predetermined constant added to the expected value is provided. It may be displayed on the screen. At this time, when a constant change instruction is input to the constant change instruction input section, the subject suggestive independent word selection step is restarted, and the number of co-occurrence and the expected value plus the changed constant are compared and the document is compared. Reselect an independent word that suggests the subject of. Then, the co-occurrence network is recreated based on the reselected independent word and displayed on the screen.

Next, the document database display device of the present invention will be described. The document database display device includes a co-occurrence table, a co-occurrence probability calculation unit, a co-occurrence count expected value calculation unit, a subject suggestion independent word group selection unit, a co-occurrence network generation unit, a keyword input unit, a document search unit, a co-occurrence network designation. Means and text reference means.

The co-occurrence table records the co-occurrence word for each independent word and the number of times of co-occurrence for each independent word extracted from a given document. The co-occurrence probability calculating means calculates a co-occurrence probability representing the strength of the co-occurrence relationship based on the co-occurrence table.

The co-occurrence count expected value calculation means calculates an expected value of the co-occurrence count based on the co-occurrence table. The subject suggesting independent word group selecting means compares the expected value and the number of times of co-occurrence and suggests the subject of the document (an independent word as a co-occurring word,
Select an independent word as a co-occurrence word).

The co-occurrence network generating means sets the magnitude of the co-occurrence probability between a character string representing an independent word selected by the subject suggestive independent word group selecting means and a character string having a co-occurrence relationship among these character strings. A co-occurrence network is created by connecting the co-occurrence lines in a display form corresponding to the size.

The keyword input means is, for example, a keyboard, and is a device for a user to input a search expression for specifying a keyword. The document search means searches one or more co-occurrence networks generated by the co-occurrence network generation means, extracts the co-occurrence network including the keyword specified by the search expression, and displays the screen.

The co-occurrence network designating means designates an arbitrary co-occurrence network among one or more co-occurrence networks displayed on the screen by the document searching means. The original text reference means reads out the original text of the document corresponding to the co-occurrence network designated by the co-occurrence network designating means, and displays a part or all of the original text on the screen.

The co-occurrence line of the co-occurrence network is a line whose line type, direction, color, shading, length, or thickness is determined by the magnitude of the co-occurrence probability of the co-occurrence word and the co-occurrence word. Alternatively, the line may be an arrow, and the line type, direction, color, shade, length, or thickness may be arbitrarily changed by the user.

[0045]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 7 shows a conceptual diagram of the processing of the present invention. FIG. 11 is a flow chart showing the procedure of the process of the present invention.

The process of this embodiment is roughly divided into a co-occurrence database creation process (100) and a subject suggestive independent word network display process (200). In order to realize these processes, a 16-bit process, preferably 32-bit or more CPU, a main memory, a large-capacity external storage device such as a hard disk device, and an external display / output device such as a CRT or a printer are provided. Use an electronic computer system. The co-occurrence database creation process (100) and the subject suggestion independent word network display process (2)
00) means that the execution program is read from the large-capacity external storage device into the CPU and executed. Further, it is assumed that the input of the original text is input in advance by application software such as an editor or a word processor and is stored in a large-capacity external storage device in a text file format, for example.

The co-occurrence database creation process (10)
0) and subject suggestion independent word network display processing (200)
May be realized by a single electronic computer, or a co-occurrence database creating device for realizing the co-occurrence database creating process (100) and a subject suggestive independent word network display process (200). It may be configured such that it is divided into a subject suggestion independent language network display device and these devices are connected by a communication line so that digitized document data, co-occurrence database and the like can be transmitted and received by communication. Further, it goes without saying that the data exchange between the co-occurrence database creation device and the subject suggestion independent word network display device is not limited to communication, and is recorded in a recording medium such as an FD or a CD-ROM for delivery. A method etc. can be illustrated.

The co-occurrence database creation process (10
0), and subject-independent independent word network display processing (20
0) will be described. <Co-occurrence Database Creation Process (100)> The co-occurrence database creation process (100) uses the co-occurrence table (TBLn), the co-occurrence probability table (TBMn), and the expected value table (TBNn) from the digitized document data. ) And a co-occurrence database (KDB) is created, and the co-occurrence database (KDB) is recorded in a large-capacity external storage device (or main memory). This co-occurrence database creation process (100) includes an independent word extraction process (110) and a co-occurrence table creation process (12) as shown in FIG.
0), the co-occurrence probability calculation process (130), and the co-occurrence count expected value calculation process (140).

The independent word extraction process extracts independent words from the digitized document data, and calculates the total number of independent words and the number of words in the document. The co-occurrence table creation process (120)
For each independent word extracted by the independent word extraction process, a co-occurrence table (TBLn) that registers the co-occurrence word and the number of times of co-occurrence for each independent word is created.

The co-occurrence probability calculation process (130) calculates the co-occurrence probability that each independent word co-occurs with the co-occurrence word for it, for all independent words recorded in the co-occurrence table (TBLn). When there are a plurality of co-occurrence words for one independent word, the co-occurrence probability is calculated for each co-occurrence word. Furthermore, in the co-occurrence probability calculation process (130), for each independent word,
A co-occurrence probability table (TBMn) for registering a co-occurrence word corresponding thereto and a co-occurrence probability co-occurring with the co-occurrence word is created.

The co-occurrence count expected value calculation process (140) is performed for all independent words recorded in the co-occurrence table (TBLn).
The expected value of the number of times of co-occurrence with each co-occurrence word for each independent word is calculated. When there are a plurality of co-occurrence words for one independent word, the expected value of the number of co-occurrence times is calculated for each co-occurrence word. Then, for each independent word, an expected value table (TBNn) is created in which the co-occurrence word for it and the expected value of the number of times of co-occurrence with that co-occurrence word are registered.

(Subject Suggestion Independent Word Network Display Process (200)) Subject Suggestion Independent Word Network Display Process (2)
00) creates a co-occurrence network based on the co-occurrence database, outputs the created co-occurrence network to an output device such as a CRT or a printer, and retrieves the original text of the co-occurrence network and outputs it to the output device. It is a process to do. In this embodiment, a CRT is used as an output device.

Subject suggestion independent word network display processing (2
00), as shown in FIG. 7, subject suggestion independent word group selection processing (210), co-occurrence network generation processing (220),
Document search processing (230) and original text reference processing (240)
Can be divided into

The subject suggestion independent word group selection process (210)
The co-occurrence database (KDB) corresponding to the document data designated by the user, that is, the co-occurrence table, the co-occurrence probability table, and the expected value table are read from the large-capacity external storage device. Then, for all independent words, the independent words and the relations between the independent words and the co-occurrence words corresponding thereto (that is, the relation between the expected co-occurrence count value and the actual co-occurrence count) satisfy the above equation (4). A combination with a co-occurrence word for is determined, and a table for registering these combinations is created. Normally, the constant α in the equation (4) has an initial value (1.5)
Is set.

The co-occurrence network generation process (220)
A co-occurrence network corresponding to each document is created by referring to the table created by the subject suggestion independent word group selection process (210). The created co-occurrence network is displayed on the CRT screen. This co-occurrence network is shown in FIG.
As shown in, the character string indicating an independent word and the co-occurrence line connecting the character strings indicating the independent words having a co-occurrence relationship are formed. The co-occurrence line has its line type, color, and
The shade, length, or thickness is different. When determining the line type, color, shading, length, or thickness of the link, refer to the co-occurrence probability in the co-occurrence probability table and, depending on the size of the co-occurrence probability, the link line type, color, and length. , Or determine the thickness. In addition, as a method to differentiate the strength of co-occurrence relationship,
It is also possible to exemplify a method in which two independent words having a co-occurrence relationship and a co-occurrence line connecting these independent words are displayed by a three-dimensional expression, and the expression form of the three-dimensional expression is changed according to the magnitude of the co-occurrence probability. Further, when the co-occurrence line is represented by a shade depending on the strength of the co-occurrence relationship, it is possible to make only those having a co-occurrence relationship stronger than a certain degree become substantially visible to the user by the user's selection. , The co-occurrence relationship can be made substantially visible to the user.

In the co-occurrence network generation processing (220), when a plurality of documents are designated by the user,
A co-occurrence network is generated for each document. When displaying these co-occurrence networks on the screen, the user can select to display all documents displayed side by side on one screen or display one document at a time on one screen. Has become. It is also possible to change the expression level (co-occurrence level) of the co-occurrence network by changing the value of the constant α in the expression (4) and reselecting the subject suggestive independent word group.

The document search process (230) searches one or more co-occurrence networks generated by the co-occurrence network generation process (220) when the user inputs a keyword from a keyboard or the like. , The co-occurrence network including the keyword input by the user is extracted. Here, when there are a plurality of co-occurrence networks including a keyword, those co-occurrence networks are displayed as all documents or one document at a time, as selected by the user.

In the original text reference process (240), the original text of the document designated by the user is read out from the document data of the mass storage device and displayed on the screen. As a procedure for designating the original sentence, the display coordinates of the co-occurrence network of the document desired by the user among the one or more co-occurrence networks displayed on the screen by the document retrieval processing (230) are designated by the position designating means. You can specify it. The position specification means
For example, it is a pointing device such as a mouse.
If the user specifies the display position of one or more arbitrary independent words on the co-occurrence network with a pointing device before the original sentence is specified, the independent word is the second keyword of the present invention. Is recognized as And
When displaying the original sentence, the sentence including the second keyword and the sentences before and after the sentence are picked up and displayed from the original sentence. At this time, the second keyword or the sentence including the second keyword is emphasized or distinguished by a method such as hatching display or reverse display. Furthermore, if there are multiple sentences containing the keyword in the original sentence, the user can specify to scroll or jump to the previous portion or sentence of the sentence currently containing the keyword, or You can scroll or jump to parts or sentences.

The operation of this embodiment will be described below. First, the flow of the co-occurrence database creation process (100) will be described taking the "Mermaid Princess" of the Andersen fairy tale as an example.

The text of "Mermaid Princess" is text-input by an editor or a word processor and is stored in the large-capacity external storage device as document data. This "Mermaid Princess"
The total number of sentences is 687, and these sentences are input sentences 1, 2,
3 ... 687.

In the co-occurrence database creation process (100), the independent word extraction process is executed to sequentially read sentences 1 to Ns (here, Ns = 687), and the independent word w _i is extracted for each sentence. Done. This independent word extraction processing is specifically performed while the CPU refers to the independent word / adjunct word dictionary (DIC) set on the large-capacity external storage device. This independent word / adjunct word dictionary (DIC) stores the stem and utilization of independent words, connection information of auxiliary words that may accompany the independent words, and the like. Further, in the independent word / adjunct word dictionary (DIC), an independent word excluded from the subject of the independent word of the present invention, for example, an independent word composed of a conjunction, a conjunction, an adverb, a formal noun, and a part of speech such as a prefix. Independent words that are unlikely to be the subject of the document are also stored.

In this way, as a result of the independent word extraction processing, the document "mermaid", "mermaid", "princess",
Words such as “prince”, “kiss”, “do” and “young” were extracted as independent words w _i .

Next, the co-occurrence table creation process (120)
Is executed, the co-occurrence word c _j and the co-occurrence number M (c _j / w _i ) are calculated for each of the independent words w _i extracted by the independent word extraction process (110). Then, for each independent word w _i , the co-occurrence word c _j and its co-occurrence count M (c _j
/ _{W i)} and co-occurrence table for registering (TBL1) is generated. Here, a specific example of the co-occurrence table (TBL1) is shown in FIG.

As shown in the figure, for example, the independent word "mermaid" and its co-occurrence word "hime" co-occur in the document 47 times, while the independent word "hime" and its co-occurrence word co-occur. The occurrence word "mermaid" co-occurs in the document 50 times. The co-occurrence table (TBL created in this way
1) is stored in a large capacity external storage device.

Next, the co-occurrence probability calculation process (130) is executed, and the co-occurrence probability P (c _j / w) is based on the co-occurrence number M (c _j / w _i ) registered in the co-occurrence table (TBL1). w _i )
Is calculated. The co-occurrence probability P (c _j / w _i ) is calculated based on the above-mentioned equation (1). For example, the number of occurrences N _wi of the independent word “mermaid” in the document is 74 times. Therefore, the co-occurrence probability P (c _j / w _i ) that the independent word “mermaid” co-occurs with the co-occurrence word “princess” is as follows.

P (c _j / w _i ) = M (c _j / w _i ) / N
_wi = 47/74 = 0.635135 The calculated co-occurrence probability P (c _j / w _i ) is registered in the co-occurrence probability table (TBMn) together with the independent word w _i and the corresponding co-occurrence word c _j. .

In FIG. 9, the co-occurrence probability P (c _j / w _i ) is registered in the co-occurrence probability table (TBMn) of the same format as the co-occurrence table (TBL1). (C _j / w _i ) may be stored in a large-capacity external storage device or the like with a certain relation with the co-occurrence table (TBL1). Further, it may be stored in a single table together with an expected value E (c _j / w _i ) of the number of times of co-occurrence described later. Furthermore, the co-occurrence count M (c _j /
w _i ), the co-occurrence probability P (c _j / w _i ) and the co-occurrence count expected value E (c _j / w _i ) described later may be stored in association with each other.

Next, the co-occurrence count expected value calculation process (140)
Is executed, and the expected value E (c _j / w _i ) of the number of times of co-occurrence is calculated based on the above equation (2). As a result, the co-occurrence count expected value E (c _j / w _i ) at which the independent word “mermaid” co-occurs with the co-occurrence word “princess” becomes 16.575816.
Expected value E (c _j / co-occurrence number) calculated as described above
w _i ) is registered in the expected value table (TBNn) together with the independent word w _j and the co-occurrence word c _{j corresponding} thereto, and this expected value table (TBNn) is stored in the co-occurrence database KDB of the large capacity external device.

Through the above processing, the co-occurrence database (K
DB) creation processing is completed. In FIG. 7, the co-occurrence table (TBL1, 2, ... N) and the co-occurrence probability P (c _j / w)
_i )) co-occurrence probability table (TBM1, 2, ...
N) and the expected value table (TBN1, 2, ..., N) that registers the expected number of co-occurrence values E (c _j / w _i ) form a co-occurrence database (KDB).

Next, the flow of the subject suggestive independent word network display processing (200) will be described. In the subject suggestion independent word network display process (200), first, a subject suggestion independent word group selection process (210) is executed. In the subject suggestion independent word group selection processing (210), the co-occurrence database corresponding to the document designated by the user (for example, the fairy tale "Mermaid Princess" in this case) is read from the co-occurrence database group of the large-capacity external storage device. Be done. Then, as shown in FIG. 11, the above equation (4) is examined for each of all independent words w _i and all co-occurrence words c _j . In addition, here, 1.5 is set as an initial value for the constant α.

Then, a table for registering the combination of the independent word w _i satisfying the above-mentioned expression (4) and the co-occurrence word c _{j for} it is created. Next, the co-occurrence network generation process (220) is executed. In this co-occurrence network generation process (220), the subject suggestion independent word group selection process (2
The combination of the independent word w _i and the corresponding co-occurrence word c _j is read from the table created in 10), and the link from the independent word w _i to the corresponding co-occurrence word c _j is stored in the large-capacity external storage device. Is passed to the display module that has been created. The display module is a character string representing the independent word w _i ,
A character string representing the co-occurrence word c _j , and an independent word w
A co-occurrence network consisting of co-occurrence lines connecting _i and the co-occurrence word c _{j corresponding} thereto is formed on the screen. At this time,
The thickness of the co-occurrence line is determined based on the level table (stored in the large-capacity external storage device) shown in Table 1 above. still,
The strength of the co-occurrence relation is determined not by the thickness of the line connecting the independent word w _i and the corresponding co-occurrence word c _j , but by the line type, color, shade, length, or three-dimensional display. Good.

A user who refers to the co-occurrence network output from a CRT or a printer can change the co-occurrence level. In this case, the value of the constant α in the above equation (4) is changed, so the subject suggestive independent word group selection process (210) and the co-occurrence network generation process (220) are executed again. In the subject suggestion independent word group selection process (210), combinations of the independent word w _i and the co-occurrence word c _{j corresponding} thereto that do not satisfy the expression (4) in which the value of the constant α is changed are selected, and these combinations are registered. The created table is recreated. Then, in the co-occurrence network generation process (220), a new co-occurrence network is created based on the recreated table and is output to the CRT.

These subject suggestion independent word group selection processing (21
0) and the co-occurrence network generation process (220) are executed for all the selected documents when the user selects a plurality of documents. A plurality of co-occurrence networks created by this may be displayed on one screen at the same time, or only one co-occurrence network may be displayed on one screen and each screen may be switched by screen switching or page switching. The co-occurrence network may be displayed.

In the document search process (230), when the user inputs an arbitrary keyword (independent word) from an input device such as a keyboard, the input keyword and the independent word included in the co-occurrence network of each document are displayed. The co-occurrence networks including the independent words that are compared and match the input keyword are extracted. The extracted co-occurrence network is a CRT
Is displayed on the screen. At this time, on the co-occurrence network, the independent word that matches the keyword input by the user is displayed so as to be emphasized or distinguished from other independent words. For example, the display color and size of an independent word that matches a keyword,
The typeface is made different from other independent words, and the independent words are shaded, highlighted, or surrounded by a square.

Since a plurality of keywords can be input, when a plurality of keywords are input, the co-occurrence network including all the keywords is determined.

By the above processing, the user can
By referring to the co-occurrence network output to T or a printer, the subject of each document can be visually recognized.
Then, the user selects a document for which the original sentence is desired to be referenced. To select a document, the user double-clicks on the screen the display coordinates of the co-occurrence network of the document desired by the user with a mouse as a position designation means. Thereby, the original sentence reference process (240) is executed, the co-occurrence network of the display coordinates designated by the mouse is discriminated, and the original sentence of the document corresponding to this co-occurrence network is read from the large capacity external storage device. And from the read original text,
The sentence including the keyword input by the user and the sentences before and after the sentence are picked up and displayed on the CRT.

Next, the operation procedure when the subject suggestive independent word network display processing (200) is realized by an electronic computer equipped with a CRT and a hard disk device as a large-capacity external storage device will be described with reference to FIGS. 13 to 31. explain.

FIG. 13 is an initial screen displayed on the CRT when the subject suggestive independent word network display processing (200) is started. This initial screen is provided with a menu bar "File", "Display", and "Help". The user selects "File" from these menu bars. When the menu bar “file” is selected, a pull-down menu of “file” is displayed as shown in FIG. In this pull-down menu,
"New", "Open" and "Exit"
And are set. "New creation" is selected when a new document is read, that is, when a co-occurrence network of a new document is created. "Open" is selected when displaying the already read document or the co-occurrence network of the document. "End of application" is the subject suggestion independent word network display process (20
Select to end 0).

When the user selects "New" from the pull-down menu, the "New" dialog box as shown in FIG. 15 is displayed. In this dialog box, the location where the document file is stored in the hard disk device "bunsyo" (default)
And the number of documents stored in the document file "4"
And are displayed. In addition, the dialog box
"Display all documents" button, "Display one document at a time" button,
"Cancel" button is set. The "Display all documents" button is a button to select when displaying multiple co-occurrence networks side by side on one screen, and the "Display one document at a time" button displays multiple co-occurrence networks on one screen in an overlapping manner. This is the button to select when displaying.
The "Cancel" button is a button to be selected when ending the new creation process.

When the display coordinates of the "display all documents" button is designated by the pointing device, the subject suggestion independent word group selection process (210) and the co-occurrence network generation process (22) of the subject suggestion independent word network display process (200).
0) and are executed to create a co-occurrence network for each document in the document file. Then, the created co-occurrence networks are displayed side by side on one screen as shown in FIG. On this screen, "Edit" and "Window" are added in addition to the menu files "File", "Display" and "Help".

Here, when the user selects "Edit" from the menu bar, a pull-down menu "Search" is displayed as shown in FIG. Subsequently, when the user selects the "search" pull-down menu, the "search" dialog box is displayed as shown in FIG. This dialog box is provided with an area for inputting a search character string (independent word) (up to 3 words can be input in the drawing), a "search execution" button, and a "cancel" button. The search character string is input from an input device such as a keyboard (see FIG. 19). When the input of the search character string is completed, the user selects the "search execution" button.

When the "search execution" button is selected, in the subject suggestion independent word network display processing (200), the document search processing (230) is executed and a co-occurrence network including all the input search character strings is displayed. To be extracted. And
On the extracted co-occurrence network, the independent word w _i that matches the search character string is displayed in a different format, such as a color, size, typeface, etc., different from other independent words (in FIG. 19, a boxed box is displayed. ). As a result, the user can visually understand the subject of each document, and can easily select the document for which the original sentence is desired to be referred to.

When the user refers to the original sentence, the user double-clicks the display coordinate of the co-occurrence network or the display coordinate of the display window of the co-occurrence network corresponding to the original sentence. By this operation, a dialog box of the original sentence as shown in FIG. 20 is displayed on the screen. In this dialog box, a part of the original sentence of the document corresponding to the selected co-occurrence network is displayed. A part of this original sentence displays the sentence that appears earliest and the sentences before and after it among the sentences that include the search character string in the original sentence, and the sentence that includes the search character string is shaded. .

Further, in the dialog box, a search character string as a keyword input at the time of document search processing,
A numerical value indicating how many sentences including these search character strings are included in the original sentence and a numerical value indicating which sentence among the sentences including the search character string the currently displayed sentence (see FIG. 2).
0, number of cases), "Previous" button, "Next" button,
An "OK" button is displayed.

The "previous" button is a button selected when referring to the sentence before the currently displayed sentence among the sentences including the search character string. When the display coordinates of the "previous" button are designated by the pointing device, the display screen scrolls or jumps to the sentence before the currently displayed sentence among the sentences including the search character string.

The "next" button is a button selected when referring to the sentence next to the currently displayed sentence among the sentences including the search character string. When the display coordinates of the "next" button are designated by pointing, the display screen scrolls or jumps to the sentence next to the currently displayed sentence among the sentences including the search character string.

Even when the display screen is scrolled by the "previous" button or the "next" button, the sentence including the search character string is displayed in a shaded manner. The “OK” button is a button selected when closing the dialog box.

If such a procedure is performed for each document,
You can refer to the original text of all documents. As a result, the user can distinguish the necessary document from the unnecessary document.

FIG. 21 shows a screen after the document 4 is saved or discarded. In this embodiment, when the document 4 is saved or discarded, the display window of the co-occurrence network corresponding to the document 4 is deleted from the screen.

When the user selects the "Display 1 document at a time" button on the screen shown in FIG. 15 described above, the display windows of the co-occurrence networks overlap as shown in FIG. Only the originating network is displayed. On this screen, in addition to the menu bar "File", "View", "Help",
"Edit" and "window" are added.

Here, in order to execute the document search process (230), as in the case of "display all documents" described above, the menu bar "Edit" is selected and the pull-down menu is read out. Select "Search" from. When "Search" is selected, a "Search" dialog box is displayed as shown in FIG. The configuration of this dialog box is the same as in the case of "display all documents".

When the display coordinates of the "search execution" button are specified by the pointing device after the search character string as the keyword is input by the keyboard,
The co-occurrence network including all the input search strings is extracted, and the display window of the co-occurrence network is moved to the foreground. At this point, the search process is temporarily interrupted (see FIG. 24).

In the co-occurrence network moved to the foreground, the independent word that matches the search character string is displayed in a different format, such as a different color and typeface, from the other independent words in the co-occurrence network (FIG. 24). Display the square box inside).

When the display coordinates of the "search execution" button are designated again by the pointing device, the search process is restarted. When the corresponding co-occurrence network is found, the display window of the co-occurrence network is moved to the foreground instead of the display window of the co-occurrence network previously displayed (see FIGS. 25 and 26).

Further, when referring to the original sentence of the document corresponding to the foreground co-occurrence network, a keyword (second keyword) different from the keyword (first keyword) input in the document search processing is used. And a procedure for referring to the original sentence and a procedure for referring to the original sentence without inputting the second keyword.

First, the procedure for referring to the original sentence without selecting the second keyword is, for example, when referring to the original sentence corresponding to the co-occurrence network shown in FIG. Double click
As a result, the original text dialog box is displayed on the screen. The displayed original sentence is a sentence including the first keyword and sentences before and after the sentence.

The procedure for selecting the second keyword and referring to the original sentence is, for example, by clicking the display coordinates of the desired independent word on the co-occurrence network shown in FIG. 27 with the mouse. At this time, although not shown, the selected independent word is displayed in a format different from other independent words. Next, when the display coordinates of the co-occurrence network (display coordinates of the display window of the co-occurrence network) are double-clicked with the mouse, a dialog box of the original sentence is displayed as shown in FIG. In this dialog box, among the original sentences, the sentence containing the second keyword and the sentences before and after it are displayed. Of these, the sentence including the second keyword is shaded.

Further, in the dialog box, the character string of the independent word selected as the second keyword, the numerical value indicating how many sentences including the second keyword are included in the original document, and the current display are displayed. A numerical value indicating the number of the sentence in which the existing sentence includes the search character string in the original sentence,
A "previous" button, a "next" button, and an "OK" button are displayed. The respective functions of the "previous" button, the "next" button, and the "OK" button are the same as in the case of "display all documents".

When the document corresponding to the co-occurrence network located in the foreground is discarded, the display window of the co-occurrence network is erased and the display window of the next co-occurrence network is moved to the foreground. .

Next, a procedure for changing the expression level (co-occurrence level) of the co-occurrence network will be described. In both cases of "display all documents" and "display one document at a time", "+" button and "-" are displayed in the display window of the co-occurrence network.
Buttons are provided, and when the display coordinates of one of the buttons is designated by the pointing device, the expression level is changed. For example, in the case where the co-occurrence networks are overlapped and displayed as shown in FIG. 29, the user must increase the expression level (decrease the co-occurrence level) by
The display coordinates of the "+" button are specified by the pointing device. When the display coordinates of the "+" button are designated, the subject suggestion independent word group selection process (210) and the co-occurrence network generation process (220) of the subject suggestion independent word network display process (200) are executed again. In the subject suggestive independent word group selection process (210), the value of the constant α in the expression (4) is reduced, and the independent word w _i and the co-occurrence word c corresponding thereto satisfying the expression (4) in which the value of the constant α is decreased. Select the combination of _j . Then, the co-occurrence network generation process (22
In 0), the co-occurrence network is recreated based on the combination of the newly selected independent word w _i and the corresponding co-occurrence word c _j . The recreated co-occurrence network is C
It is displayed on the RT (see FIG. 30).

On the other hand, in order to lower the expression level (raise the co-occurrence level), the user specifies the display coordinates of the "-" button with the pointing device. When the display coordinates of the "-" button are designated, the subject suggestion independent word group selection process (2) of the subject suggestion independent word network display process (200).
10) and the co-occurrence network generation process (220) are executed again. In the subject suggestion independent word group selection process (210),
The value of the constant α in the equation (4) is increased, and a combination of the independent word w _i and the co-occurrence word c _{j corresponding} thereto that satisfy the equation (4) in which the value of the constant α is increased is selected. Then, in the co-occurrence network generation processing (220), the co-occurrence network is re-created based on the combination of the newly selected independent word w _i and the co-occurrence word c _{j corresponding} thereto. The recreated co-occurrence network is displayed on the CRT (see FIG. 31).

As described above, according to this embodiment, the user is
You can visually understand the subject of each document without reading the original text. Also, the operation performed to understand the subject of each document is simple.

As a method of expressing the co-occurrence network, as shown in FIG. 32, the strength of the co-occurrence relation between the independent words may be differentiated by the line type of the straight line connecting the independent words. . In the figure, the co-occurrence lines are classified into three types: thick solid lines, thin solid lines, and dotted lines. As a method of determining the line type, three constants A, B, and C (A>B> C) are preset.
Is established, and the independent words having the co-occurrence relationship in which the difference (left-side and right-side) is larger than the constant A are satisfied by a thick solid line, and the above-mentioned (4) is satisfied and the difference is The independent words having a co-occurrence relation that is equal to or greater than the constant B and is equal to or less than the constant A are connected by a thin solid line. A method of tying can be exemplified. The linetype of the co-occurrence line connecting each independent word is
When the expression level (co-occurrence level) is changed by the "+" button and the "-" button, it is changed in accordance with this change.

Further, as a co-occurrence network expression method, as shown in FIGS. 33 to 36, the strength of the co-occurrence relationship between independent words is differentiated by the gradation (shading) of independent words and co-occurring lines. Can be illustrated. in this case,
In the co-occurrence network generation process, a co-occurrence network with a high expression level (weak co-occurrence level) is generated beforehand (see FIG. 33).
Is generated. At the time of generation, the co-occurrence levels between independent words that have a co-occurrence relationship are also classified into several levels according to their size. For example, when classifying co-occurrence levels into three levels, three constants A, B, and C (A>B> C) are determined in advance, and (condition 1) the above-mentioned expression (4) is satisfied and the difference is Between independent words having a co-occurrence relationship in which (left side-right side) is larger than a constant A, (condition 2) self-supporting relationship in which the above expression (4) is satisfied and the difference is a constant B or more and a constant A or less. Words are classified into (condition 3) three independent word terms that satisfy the above-mentioned expression (4) and have a difference of less than a constant C in a co-occurrence relationship.
Then, when the co-occurrence network is displayed on the screen, for example, as shown in FIG. 34, only the independent word and the co-occurrence line satisfying the above (condition 1) are displayed. In the screen of Fig. 34, the part with a dark gradation ("hair" and "hair", "transparent" in the figure)
And "passed") have reached the user's visible level.

When the user selects the "+" button or "-" button on the screen, the gradation is changed. For example, if the "+" button on the screen is selected,
As shown in FIG. 3, the color of the independent word that satisfies the above (condition 2) and the co-occurrence line gradually darken, and finally, as shown in FIG. 36, the independence that satisfies (condition 1) It is as thick as words and co-occurrence lines. At this time, the user has (condition 1)
In addition to the independent words and co-occurrence lines that satisfy (Condition 2), independent words and co-occurrence lines that satisfy (Condition 2) are visible.

When the independent word and co-occurrence line satisfying the above (condition 1) and the independent word and co-occurrence line satisfying the above (condition 2) are displayed on the screen, the "-" button on the screen is pressed. When selected, the colors of the independent words and the co-occurrence lines that satisfy (Condition 2) gradually become lighter, and finally become thinner to a level that cannot be recognized by human eyesight. At this time, the user can see only the independent word and the co-occurrence line that satisfy (Condition 1).

In addition to the line types of co-occurrence lines and gradation,
A method of differentiating by the color of co-occurrence line, independent word, or three-dimensional expression may be used. <Other Embodiments> When the present invention is implemented by a co-occurrence database creation device and a subject suggestion independent word network display device, the co-occurrence database creation device is arranged on the server side and the subject suggestion independent word network display device is provided. It is conceivable to arrange the client on the client side and connect these devices with a communication line. When this method is adopted, it is preferable that the co-occurrence database generation device performs the subject suggestive independent word group selection process and the co-occurrence network generation process to generate the co-occurrence network in advance. Then, the subject suggestive independent word network display device transmits a co-occurrence network transmission request including an arbitrary keyword to the co-occurrence database creation device. The co-occurrence database creation device searches a co-occurrence network group created in advance to extract a co-occurrence network including an arbitrary keyword,
Sending to the suggestion of independent subject network display device.

The subject suggestion independent language network display device is
When the co-occurrence database is received from the co-occurrence database creating device, the co-occurrence database is displayed on the screen by a display device such as a CRT.

When the user referring to the co-occurrence network on the screen inputs a reference request for the original sentence (the display coordinates of the co-occurrence network is designated by the position designating means), the subject suggestive independent word network display device is designated by the user. The identification information of the co-occurrence network is transmitted to the co-occurrence database creation device.

Upon receiving the identification information from the subject suggestive independent word network display device, the co-occurrence database creating device reads out the original sentence of the document corresponding to the co-occurrence network identified by the identifying information and displays the subject suggestive independent word network. Send to the device.

In the subject suggestive independent word network display device that has received the original sentence, the original sentence is displayed on the screen by the same procedure as in the above-described embodiment. When this method is adopted, when a user who refers to the co-occurrence network desires to change the expression level, the co-occurrence database is recreated by the co-occurrence database creation device, and the re-created co-occurrence network is re-subjected to the subject. Although it may be sent again to the suggestion independent word network display device, a co-occurrence network with a high expression level should be created in advance with the co-occurrence database creation device, and the level of the co-occurrence line between each independent word in the co-occurrence network should be You may make it set an identifier. The level identifier mentioned here is information for identifying the strength of the co-occurrence relationship between the independent words.

For example, when classifying the strength of the co-occurrence relationship into three levels, three constants A, B, C (A>
B> C), (condition 1) the above-mentioned (3) (or (4)) is satisfied, and the difference (left side-right side) is a co-occurrence relation larger than the constant A, between independent words, ( Condition 2) Between independent words satisfying the above-mentioned expression (3) (or expression (4)) and having a difference of not less than a constant B and having a constant A or less, (condition 3) the above-mentioned expression (3) (Or (4)
(3) and the difference is less than the constant C and is classified into three independent words having a co-occurrence relationship.

A level identifier "1" is given between independent words that satisfy the above condition 1, a level identifier "2" is given between independent words that satisfy the above condition 2, and a level identifier is given between independent words that satisfy the above condition 3. The identifier “3” is added.

The co-occurrence network creation device stores the co-occurrence network to which the level identifier is added in the mass storage device. Then, when a co-occurrence network transmission request including an arbitrary keyword is transmitted from the subject suggestive independent word network display device to the co-occurrence network creation device, the co-occurrence network creation device is stored in the mass storage device. Then, the co-occurrence network including the arbitrary keyword and its level identifier are extracted. Then, the co-occurrence network creation device transmits the level identifier together with the co-occurrence network to the subject suggestion independent word network display device.

The theme suggestive independent word network display device is
First, the space between independent words to which the level identifier “1” is added and the space between independent words to which the level identifier “2” is added are displayed on the screen. Then, when the user specifies to increase the expression level (decrease the co-occurrence level), the subject suggestive independent word network display device displays between the independent words to which the level identifier “1” is added and the level identifier “2”. The space between the independent words to which "" is added and the space between the independent words to which the level identifier "3" is added are displayed on the screen.

On the other hand, when the user designates to lower the expression level (increase the co-occurrence level), the subject suggestive independent word network display device screens only the independent words to which the level identifier "1" is added. indicate.

[0117]

According to the present invention, the co-occurrence relationship is used to show the semantic connection between words, so that the operator can visually understand the document subject.

By providing a constant for increasing or decreasing the number of links, it is possible to create a general-purpose database and extract a co-occurrence network regardless of the type of document.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing a specific example of a co-occurrence table.

FIG. 2 is an explanatory diagram showing a specific example of a co-occurrence table.

[FIG. 3] Co-occurrence count M (c _j / w _i ) and its expected value E
Graph diagram showing the relationship with (c _j / w _i ) at α = 1.5

FIG. 4 is the co-occurrence count M (c _j / w _i ) and its expected value E
Graph diagram showing the relationship with (c _j / w _i ) at α = 2.5

FIG. 5: Co-occurrence count M (c _j / w _i ) and its expected value E
Graph showing the relationship with (c _j / w _i ) at α = 0.5

FIG. 6 is an explanatory diagram showing an example of an expression form of a co-occurrence network.

FIG. 7 is a block diagram showing the concept of processing of the present invention.

FIG. 8 is an explanatory diagram showing a co-occurrence table according to the embodiment.

FIG. 9 is an explanatory diagram showing a calculation result of a co-occurrence probability according to the embodiment.

FIG. 10 is an explanatory diagram showing a calculation result of an expected value of the number of times of co-occurrence according to the embodiment.

FIG. 11 is a flowchart showing a co-occurrence database creation process and a subject suggestive independent word network display process according to the embodiment.

FIG. 12 is an explanatory diagram showing a display example of the co-occurrence network according to the embodiment.

FIG. 13 is an initial screen when the subject suggestive independent word network display process is executed.

FIG. 14 is an operation screen (1) when the subject suggestion independent word network display process is executed.

FIG. 15 is an operation screen (2) when the subject suggestive independent word network display process is executed.

FIG. 16: Display screen of co-occurrence network

FIG. 17: Operation screen for document search processing (1)

FIG. 18: Operation screen for document search processing (2)

FIG. 19: Operation screen for document search processing (3)

[Figure 20] Original text display screen

FIG. 21: Co-occurrence network display screen after storing or discarding the original text

FIG. 22: Display screen of co-occurrence network

FIG. 23: Operation screen for document search processing (4)

FIG. 24: Operation screen for document search processing (5)

FIG. 25: Operation screen for document search processing (6)

FIG. 26: Operation screen for document search processing (7)

FIG. 27: Operation screen for original text reference processing

Fig. 28 Display screen of original text

[Fig. 29] Operation screen for changing the expression level of the co-occurrence network

FIG. 30: Display screen after changing the co-occurrence network expression level (1)

FIG. 31: Display screen after changing the expression level of the co-occurrence network (2)

FIG. 32 is a diagram (1) showing another display format of the co-occurrence network.

FIG. 33 is a diagram (2) showing another display format of the co-occurrence network.

FIG. 34 is a diagram (3) showing another display format of the co-occurrence network.

FIG. 35 is a diagram (4) showing another display format of the co-occurrence network.

FIG. 36 is a diagram (5) showing another display format of the co-occurrence network.

[Explanation of symbols]

 DIC-independent word-adjunct dictionary TBL1-n-Co-occurrence table TBM1-n-Co-occurrence probability table TBN1-n-Expected value table KDB-Co-occurrence database 100-Co-occurrence database creation process 110-・ Independent word extraction process 120 ・ ・ Co-occurrence table creation 130 ・ ・ Co-occurrence probability calculation process 140 ・ ・ Co-occurrence frequency expected value calculation process 200 ・ ・ Thematic suggestion independent word network display process 210 ・ ・ Thematic suggestion independent word group selection process 220 ... Co-occurrence network generation processing 230 ... Document search processing 240 ... Original text reference processing

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] March 13, 1996

[Procedure Amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0004

[Correction method] Change

[Correction content]

[0004] The former, "in general, what you want to emphasize in the document are those mentioned repeatedly, frequently there is a high possibility that indicate the subject matter of the document words that appear" is based on the theory that. However, in the independent words, "and",
"That", "this", "early", "(such a thing)",
There are words that appear as linguistic functions (connectives, adnominals, adverbs, formal nouns, prefixes, suffixes, part of demonstrative pronouns, etc.) such as "ka (yowai)" and "(tsuyaya) poi". . Therefore, it is not possible to determine the important word only by the appearance frequency. It is necessary to exclude such language function words and extract a word that has a meaning independently and represents a feature of a document. For that purpose, first, an important word list having a large vocabulary is prepared in advance, and then the appearance frequency of a single word in a document is calculated to determine a keyword. In determining this keyword, the result of linguistic analysis described below may be used. Then, a summary is generated based on the words that appear frequently in this keyword group.

[Procedure Amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0013

[Correction method] Change

[Correction content]

FIG. 2 shows an example of creating a co-occurrence table based on the sentence "I go to a cooking school to become a cooking teacher." Next, a co-occurrence probability calculating step for calculating a co-occurrence probability representing the strength of the co-occurrence relationship is executed based on the co-occurrence table. Here, the co-occurrence probability is
It is the probability P (cj / wi) that the co-occurrence word cj for w i co-occurs with wi when a certain independent word wi appears, and can be calculated by the following equation (1).

[Procedure 3]

[Document name to be amended] Statement

[Name of item to be corrected] 0032

[Correction method] Change

[Correction content]

When the co-occurrence network is displayed on the screen,
For example, the independent word that matches the keyword may be emphasized or displayed separately from others. The method emphasize or discriminate can illustrate how to display a different color, font, or size, etc. The keywords and match independent words and other independent word.

[Procedure amendment 4]

[Document name to be amended] Statement

[Correction target item name] 0033

[Correction method] Change

[Correction content]

When the original sentence is displayed on the screen in the original sentence reference step, there are a method of displaying the entire original sentence on the screen and a method of displaying a part of the original sentence on the screen. When displaying a part of the original sentence on the screen, display the sentence or character string containing the keyword input in the document search step, or the sentence or character string containing the keyword and the sentence or character string before and after it. May be. At this time, it may be displayed by highlighting the sentence containing only keywords, or keyword. Examples of the method of emphasizing include a method of displaying only a keyword or a sentence including the keyword in a shaded manner, a method of displaying in reverse, a method of making a character color, a typeface, and a size different from other characters. .

[Procedure Amendment 5]

[Document name to be amended] Statement

[Correction target item name] 0035

[Correction method] Change

[Correction content]

When requesting the reference of the original text,
A method of designating the display coordinates of the co-occurrence network corresponding to the original sentence by the location designating means can be exemplified. Furthermore, since there is a possibility that there are multiple sentences containing keywords in the original sentence, one sentence of the multiple sentences is displayed on the screen, and the keyword (search formula) and the sentence containing the keyword are included in the original sentence. and a number indicating how many, and the numerical value indicating whether the statements that are currently displayed out of the sentence, including the keyword is what the second sentence, the sentence before the sentence that is currently displayed in one of the statements, including the keyword A pre-sentence display command input unit for inputting a display command and a next sentence display command input unit for inputting a display command of a sentence after the currently displayed sentence including the keyword may be displayed on the screen. .

[Procedure correction 6]

[Document name to be amended] Statement

[Correction target item name] 0036

[Correction method] Change

[Correction content]

At this time, the user uses a numerical value indicating how many sentences including the keyword are in the original sentence, and a numerical value indicating the number of the sentence currently displayed among the sentences including the keyword. By referring to, it is possible to recognize that the sentence including the keyword exists before or after the currently displayed sentence. Then, when the user inputs a preamble display command to the preamble display command input unit, the sentence before the currently displayed sentence among the sentences including the keyword is displayed on the screen and the sentence including the keyword is currently displayed. The number that indicates which sentence the current sentence is is 1
Is decremented. Further, if the user inputs the next sentence display command to the next sentence display command input unit, with the following statement statement currently displayed among the sentences containing the keywords is displayed on the screen, of the sentence containing the keywords The numerical value indicating the order of the currently displayed sentence is incremented by one.

[Procedure Amendment 7]

[Document name to be amended] Statement

[Correction target item name] 0040

[Correction method] Change

[Correction content]

The co-occurrence count expected value calculation means calculates an expected value of the co-occurrence count based on the co-occurrence table. The subject suggesting independent word group selecting means compares the expected value with the number of times of co-occurrence to select an independent word group suggesting the subject of the document.

[Procedure Amendment 8]

[Document name to be amended] Statement

[Correction target item name] 0041

[Correction method] Change

[Correction content]

The co-occurrence network generating means includes a character string representing each independent word of the independent word group selected by the subject suggestive independent word group selecting means, and a character having a co-occurrence relationship among the character strings. A co-occurrence network is generated by connecting the columns with a co-occurrence line having a display form according to the magnitude of the co-occurrence probability.

[Procedure Amendment 9]

[Document name to be amended] Statement

[Correction target item name] 0042

[Correction method] Change

[Correction content]

The keyword input means is, for example, a keyboard, and is a device for a user to input a search expression for specifying a keyword. The document search means searches the one or more co-occurrence networks generated by the co-occurrence network generation means and extracts one or more co-occurrence networks including the keyword specified by the search expression. To display on the screen.

[Procedure Amendment 10]

[Document name to be amended] Statement

[Correction target item name] 0044

[Correction method] Change

[Correction content]

The co-occurrence line of the co- occurrence network is
A line or arrow whose line type, orientation, color, shade, length, or thickness is determined by the magnitude of the co-occurrence probability between certain independent words . These line types, orientation, color, shade, and length The thickness or the thickness may be arbitrarily changed by the user.

[Procedure Amendment 11]

[Document name to be amended] Statement

[Correction target item name] 0049

[Correction method] Change

[Correction content]

The independent word extraction process (110) extracts independent words from the digitized document data and calculates the total number of independent words and the number of vocabularies in the document. Co-occurrence table creation process (1
20) creates a co-occurrence table (TBLn) that registers the co-occurrence word and the number of times of co-occurrence for each independent word, for each independent word extracted by the independent word extraction process (110) .

[Procedure Amendment 12]

[Document name to be amended] Statement

[Correction target item name] 0052

[Correction method] Change

[Correction content]

(Subject Suggestion Independent Word Network Display Process (200)) Subject Suggestion Independent Word Network Display Process (2)
00) creates a co-occurrence network based on a co-occurrence database (KDB) , outputs the created co-occurrence network to an output device such as a CRT or a printer, and retrieves and outputs the original text of the co-occurrence network. This is the process of outputting to the device. In this embodiment, a CRT is used as the output device.
To use.

[Procedure Amendment 13]

[Document name to be amended] Statement

[Correction target item name] 0054

[Correction method] Change

[Correction content]

The subject suggestion independent word group selection process (210)
The co-occurrence database (KDB) corresponding to the document data designated by the user, that is, the co-occurrence table (TBLn) and the co-occurrence probability table (T
BMn) and the expected value table (TBNn) are read. Then, for all independent words, the independent words and the relations between the independent words and the co-occurrence words corresponding thereto (that is, the relation between the expected co-occurrence count value and the actual co-occurrence count) satisfy the above equation (4). A combination with a co-occurrence word for is determined, and a table for registering these combinations is created. Normal,
An initial value (1.5) is set to the constant α in the equation (4).

[Procedure Amendment 14]

[Document name to be amended] Statement

[Correction target item name] 0061

[Correction method] Change

[Correction content]

In the co-occurrence database creation process (100), the independent word extraction process (110) is executed to sequentially read sentences 1 to sentence Ns (here, Ns = 687), and the independent word wi is added to each sentence. Extraction is done. This independent word extraction processing (110) is specifically performed by the CPU referring to the independent word / adjunct word dictionary (DIC) set in the large-capacity external storage device. This independent word / adjunct word dictionary (DIC)
Stores the stem and utilization of an independent word, connection information of an accessory word that may accompany the independent word, and the like. Further, in the independent word / adjunct word dictionary (DIC), an independent word excluded from the subject of the independent word of the present invention, for example, an independent word composed of a conjunction, a conjunction, an adverb, a formal noun, and a part of speech such as a prefix. Independent words that are unlikely to be the subject of the document are also stored.

[Procedure Amendment 15]

[Document name to be amended] Statement

[Correction target item name] 0079

[Correction method] Change

[Correction content]

When the user selects "New" from the pull-down menu, the "New" dialog box as shown in FIG. 15 is displayed. In this dialog box, the location "bunsyo" (default value) where the document file is stored in the hard disk device and the number "4" of documents stored in the document file are displayed. Furthermore, a "display all documents" button, a "display one document at a time" button, and a "cancel" button are set in the dialog box. The "Display all documents" button is a button to select when displaying multiple co-occurrence networks side by side on one screen, and the "Display one document at a time" button displays multiple co-occurrence networks on one screen in an overlapping manner. a button to be selected if you <br/>. The "Cancel" button is a button to be selected when ending the new creation process.

[Procedure Amendment 16]

[Document name to be amended] Statement

[Correction target item name] 0081

[Correction method] Change

[Correction content]

Here, when the user selects "Edit" from the menu bar, a pull-down menu "Search" is displayed as shown in FIG. Subsequently, when the user selects the "search" pull-down menu, the "search" dialog box is displayed as shown in FIG. This dialog box is provided with an area for inputting a search character string (independent word) (up to 3 words can be input in the drawing), a "search execution" button, and a "cancel" button. The search character string is input from an input device such as a keyboard (see FIG. 19). When the input of the search character string is completed, the user selects the "search execution" button.

[Procedure Amendment 17]

[Document name to be amended] Statement

[Correction target item name] 0095

[Correction method] Change

[Correction content]

When referring to the original sentence of the document corresponding to the foremost co-occurrence network, the document search process (2
30) The procedure of referring to the original sentence by inputting a keyword (second keyword) different from the keyword input in (1), and referring to the original sentence without inputting the second keyword. There are procedures.

Claims (16)

[Claims] 1. A co-occurrence table for creating a co-occurrence table for recording a co-occurrence word and a co-occurrence number of the co-occurrence word for each of the extracted independent words, the independent word extraction step of extracting an independent word from a given document. A co-occurrence probability calculating step of calculating a co-occurrence probability representing the strength of a co-occurrence relationship based on the co-occurrence table; and a co-occurrence calculating an expected value of the number of co-occurrence times based on the co-occurrence table. An expected number of times calculation step, a step of selecting an independent word group having a co-occurrence relationship that can suggest the subject of the document by comparing the expected value with the number of times of co-occurrence A co-occurrence network generation step of setting a link between independent words having a co-occurrence relationship that suggests the specified subject and displaying the expression form of the link as a co-occurrence network determined based on the co-occurrence probability. How to build a document database that contains the. 2. The subject suggestion independent word group selecting step,
2. The independent word group having a co-occurrence relationship that can suggest the subject of the document is selected by comparing the number of times of co-occurrence with a value obtained by adding a predetermined constant to the expected value. How to build a document database. 3. The co-occurrence network generating step includes a character string representing each independent word of the independent word group having a co-occurrence relationship selected in the subject suggestive independent word group selecting step, and an independent word having a co-occurrence relationship. The method for constructing a document database according to claim 1, further comprising a step of displaying on screen a co-occurrence network formed from a co-occurrence line visually connecting character strings representing each. 4. The line or arrow whose line type, color, shade, direction, length, or thickness is determined by the size of the co-occurrence probability between independent words having a co-occurrence relationship. The method of constructing a document database according to claim 3, wherein 5. The one or more co-occurrence networks generated in the co-occurrence network generation step are searched to extract one or more co-occurrence networks including arbitrary keywords, A document search step of displaying the co-occurrence network on the screen, and one or two displayed on the screen in the document search step.
The document database construction according to claim 1, further comprising: an original sentence reference step of extracting an original sentence corresponding to an arbitrary co-occurrence network from one or more co-occurrence networks and displaying all or a part of the original sentence on a screen. Method. 6. The one or more co-occurrence networks generated in the co-occurrence network generation step are searched to search for one or more first keywords containing one or more first keywords.
A document retrieval step of extracting one or two or more co-occurrence networks and displaying the co-occurrence networks on the screen, and one or two displayed on the screen in the document retrieval step.
A sentence or a character string including one or two or more second keywords extracted from an original sentence corresponding to an arbitrary co-occurrence network of one or more co-occurrence networks,
2. The method for constructing a document database according to claim 1, further comprising a sentence or character string including the second keyword and an original sentence reference step of displaying the sentence or character string before and after the sentence on the screen. 7. The original text reference step, when displaying a sentence or a character string including the one or more second keywords and sentences or character strings before and after the sentence or character string on the screen, The method for constructing a document database according to claim 6, wherein sentences including the second keyword are displayed separately. 8. The document search step includes a step of displaying a search expression input section for inputting a search expression for specifying a keyword and a search execution instruction input section for inputting a search execution instruction, and the search expression When a search expression is input to the input unit and a search execution command is input to the search execution command input unit, one or more co-occurrence networks generated in the co-occurrence network generation step are searched. A step of extracting one or more co-occurrence networks including a keyword specified by the search expression, displaying the one or more co-occurrence networks extracted above on a screen, and the co-occurrence 6. A step of separately displaying keywords specified by the search expression among independent words included in a network. How to build a document database. 9. In the step of referring to an original sentence, when a sentence including the keyword is displayed on a screen, the keyword, a numerical value indicating how many sentences including the keyword are included in the original sentence, and the keyword are displayed. A numerical value indicating the number of the sentence currently displayed among the included sentences, and a pre-sentence display for inputting a display command for a sentence before the currently displayed sentence among the sentences including the keyword A command input unit and a next sentence display command input unit for inputting a display command of a sentence after the currently displayed sentence among the sentences including the keyword are displayed, and the previous sentence display command input unit is displayed. When a command is input, a sentence before the currently displayed sentence among the sentences including the keyword is displayed on the screen, and when the next sentence display command is input to the next sentence display command input unit, the keyword is changed. Sentence containing 6. The method for constructing a document database according to claim 5, wherein a sentence next to the currently displayed sentence is displayed on the screen. 10. The co-occurrence network generation step, together with the generated co-occurrence network, a step of displaying on a screen a constant change instruction input unit for inputting an instruction to change a predetermined constant added to the expected value, When a constant change instruction is input to the constant change instruction input unit, the subject suggestive independent word selection step is restarted,
Comparing the number of times of co-occurrence with a value obtained by adding a constant after the change to the expected value, and reselecting a group of independent words having a co-occurrence relationship that may suggest the subject of the document; The method for constructing a document database according to claim 2, further comprising: a step of recreating a co-occurrence network based on the independent word, and a step of displaying the re-created co-occurrence network on a screen. 11. A character string representing each independent word of an independent word group having a co-occurrence relationship that can suggest the subject of a document, and a co-occurrence line visually connecting the character strings of each independent word having a co-occurrence relationship. When a display coordinate of an arbitrary character string is designated by the position designating means and a display coordinate of the co-occurrence network is designated by the position designating means, the co-occurrence network corresponding to the co-occurring network is displayed. The original sentence of the document to be extracted is extracted, and the sentence or the character string including the character string designated by the position designating means in the original sentence, or the sentence or the character string including the character string and the sentence or the character string before and after the sentence or the character string. How to display the document database displayed on the screen. 12. The co-occurrence line is a line or arrow whose line type, direction, color, length, or thickness is determined by the magnitude of the co-occurrence probability between independent words having a co-occurrence relationship. The document database display method according to claim 11. 13. A co-occurrence table recording, for each independent word extracted from a given document, a co-occurrence word for each independent word and the number of times of co-occurrence, and a co-occurrence relation based on the co-occurrence table. A co-occurrence probability calculating means for calculating a co-occurrence probability representing strength; a co-occurrence number expected value calculating means for calculating an expected value of the co-occurrence number based on the co-occurrence table; the expected value and the co-occurrence number. And a self-supporting word group selecting means for selecting a self-supporting word group having a co-occurrence relationship that can suggest the subject of the document, and each self-supporting word group selected by the subject-suggesting independent word group selecting means. A co-occurrence network consisting of a character string representing a word and a co-occurrence line that connects between the character strings representing each independent word in the co-occurrence relationship among the character strings in a form according to the magnitude of the co-occurrence probability is defined. Co-occurrence network generation means to generate and keywords Keyword inputting means for inputting, one or more co-occurrence networks generated by the co-occurrence network generating means, and one or more or more including the keyword inputted to the inputting means. Document retrieval means for extracting a co-occurrence network and displaying it on the screen, and co-occurrence network designating means for designating an arbitrary co-occurrence network among one or more co-occurrence networks displayed on the screen by the document retrieval means. And a document database display device for reading an original sentence of a document corresponding to the co-occurrence network designated by the co-occurrence network designating unit and displaying a part or all of the original sentence on a screen. 14. A co-occurrence line composed of a character string representing each independent word of an independent word group that can suggest the subject of a document and a co-occurrence line connecting the character strings representing each independent word having a co-occurrence relationship. When the display coordinates of the co-occurrence network are specified by the position specifying means after the display coordinates of an arbitrary character string are specified by the position specifying means, the document corresponding to the co-occurrence network is displayed. Of the original text, and the sentence or character string containing the specified character string in the original text, or the sentence or character string including the character string and the sentence or character string before and after the character string and the character string A document database display device that distinguishes and displays only a sentence including the above or the character string. 15. The co-occurrence line is a line or arrow whose line type, direction, color, shade, length, or thickness is determined by the magnitude of the co-occurrence probability between independent words having a co-occurrence relationship. 15. The document database display device according to claim 14. 16. The line type, color, shade, length of the co-occurrence line,
Alternatively, the thickness is arbitrarily changed by the user.
4. A display device for the document database described in 4.