JP2003162531A - Document retrieval system and document retrieval method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To facilitate the retrieval of a document which conforms to retrieval intention. <P>SOLUTION: In a document database DB1, a plurality of documents to be used in the range of a plurality of specific fields are stored. A retrieval sentence setup means 11 makes a user specify the fields of retrieval range, as well as set up retrieval sentences. When the retrieval sentence is set by the retrieval sentence setup means 11, a retrieval processing means 12 extracts documents satisfying the retrieval conditions specified by the retrieval sentences from the document database DB1. The retrieval process means 12 decides rating criterion for rating the use value of each extracted document in the specified field by the retrieval sentence setup means 11, arranges the retrieved sentences according to the size of the rating criterion as a retrieval result, and delivers the result to a retrieval result output means 13. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document retrieval system and a document retrieval method for retrieving a document related to a desired keyword from a document database that stores documents.

[0002]

2. Description of the Related Art Generally, as a technique for retrieving a document from a document database in which documents are stored, a vocabulary as a keyword is given, and a keyword set in advance with the document or a full text of the document and a character string of the keyword are collated. The technology to do is widely adopted.

However, the keywords attached to the document and the vocabulary used in the document are not unified, and even if the concepts are almost the same, the vocabulary is often different. There is a case where the vocabulary matching with is not used for the desired document, and a proper document cannot be extracted due to so-called search omission. When a keyword is given as a free word, the possibility of omission of search can be reduced by giving multiple related vocabularies as keywords, but it is time-consuming to find the relevant vocabulary each time a document is searched. In addition, there is a problem in that an appropriate vocabulary cannot be found unless one is familiar with the field related to the document to be searched.

On the other hand, JP-A-11-126202
In the gazette, simple search conditions entered by the user are converted into search formulas registered in a table in advance, and the search conditions are expanded to increase the number of vocabularies than the input search conditions and to avoid omission of search. A technique is described which reduces the possibility and eliminates unnecessary noise by converting into a proper search expression. The form of the search expression to be expanded from the search condition is registered in the table in advance according to the search condition. That is, the search formula is fixedly determined for the search condition.

Further, in Japanese Unexamined Patent Publication No. 10-72107, when a desired document is not extracted by a given keyword, the keyword is decomposed into a plurality of vocabulary and another vocabulary related to the vocabulary included in the keyword is decomposed. A technique is disclosed that reduces the possibility of omission of search by deriving and expanding the number of vocabularies.

Further, in Japanese Unexamined Patent Publication No. 8-171569, in addition to the technique of expanding a keyword into a plurality of related vocabularies, in order to prevent the vocabulary that does not match the search intention from being included in the keywords, When a given keyword is classified into multiple categories, a vocabulary that is defined as a vocabulary is displayed, and other vocabulary in each category is presented to the user to prevent the development of vocabulary that does not match the search intention. Is listed.

[0007]

By the way, Japanese Unexamined Patent Application Publication No. H11-11
According to the technique described in Japanese Patent Laid-Open No. 126202, although it is possible to reduce the possibility of omission of search for search conditions and eliminate unnecessary noise, a search formula for search conditions is fixedly set. That is
The ambiguous search conditions that have been input are simply replaced with search conditions that are more suitable for searching. Therefore,
A search formula corresponding to the search condition must be prepared in advance, and a huge amount of labor is required to set an appropriate search formula. As a result, when a special technical term is used as a search condition, it is highly possible that the search formula corresponding to the search condition is not set, resulting in an increase in search omissions and a lot of unnecessary noise. More likely.

The technique disclosed in Japanese Patent Laid-Open No. 10-72107 only decomposes a keyword into a plurality of vocabularies and expands the number of vocabularies. Therefore, even a special technical term is relatively compatible. It is easy, but when the desired document is not extracted for the keyword, the keyword is decomposed into multiple vocabularies and the number of vocabularies is expanded. Often, the search process of is required. That is, it takes a relatively long time until the desired document is extracted. In addition, it is difficult to eliminate unnecessary noise because the vocabulary number is simply expanded.

In Japanese Unexamined Patent Publication No. 8-171569, vocabulary categories are classified into categories, and the vocabulary classified into each category is selected by the user, so that the vocabulary is expanded within the range of the categories that match the search intention. It is considered that search omission is reduced by expanding the vocabulary and unnecessary noise is relatively reduced by restricting the categories.

However, since noise cannot be eliminated even by using the technique described in this publication, a plurality of documents are often extracted according to the search condition, and finally a plurality of extracted documents are extracted. The user has to find out a desired document from among these, and the effort for finding the document that matches the search intention is still great.

The present invention has been made in view of the above reasons, and an object thereof is to associate a document to be searched with an evaluation scale for each field and sort search results in order of the size of the evaluation scale. It is to provide a document search system and a document search method in which documents that match the search intention are arranged so that the documents that match the search intention can be quickly found and the documents can be easily extracted.

[0012]

According to the invention of claim 1, a document database in which a plurality of documents used within a range of a specific plurality of fields are stored, and a user sets search texts and sets fields of the search range. Retrieval statement setting means for designating, retrieval processing means for extracting documents matching the retrieval conditions of the retrieval statement set by the retrieval statement setting means from the document database, and retrieval result output means for outputting retrieval results by the retrieval processing means The search processing means includes a function of obtaining an evaluation scale in each field of each document based on evaluation data classified by field for each document registered in the document database, and a plurality of types of evaluation methods different in the method of acquisition. A function for allowing the user to select a desired evaluation scale from the scales, and an evaluation scale of the type selected by the user for the field specified by the search sentence setting means for each extracted document. And having a function to pass the search result output unit as a search result are arranged in magnitude order of the evaluation scale with seeking.

According to a second aspect of the invention, in the first aspect of the invention, the search sentence setting means allows the user to input a primary search sentence which is a natural sentence including a keyword, and a primary search sentence. A search sentence analyzing means for extracting a keyword from the related sentence, a related term database in which the related terms for the vocabulary to be the keyword are registered for each of the fields, and the keyword extracted by the search sentence analyzing means are collated with the related term database and the search sentence analyzing means A secondary search sentence generation means having a function of generating a secondary search sentence using a related term for the extracted keyword, and a search sentence given to the search processing means from the primary search sentence and the secondary search sentence to the user. It is characterized by comprising an execution search sentence selection means for selecting.

According to a third aspect of the present invention, in the second aspect of the present invention, in the related term database, in addition to the synonyms of the vocabulary serving as a keyword, a vocabulary having a specific relevance in each field is registered as a related term. At the same time, each related term is associated with a term importance that is a measure of the utility value in each field, and the secondary search sentence generation means shows the term importance to the user together with the related term, and the related term The feature is that the desired related term is selected.

According to a fourth aspect of the present invention, in the first to third aspects of the invention, the utility value in each field of each document registered in the document database is registered as the document importance serving as the evaluation data. A document importance database is provided, and when the document importance is selected as one of the evaluation scale options in the search processing means, the document extracted by the search condition is collated with the document importance database to thereby determine the document importance. The document is arranged by using the degree as the evaluation scale.

According to a fifth aspect of the present invention, in the first to third aspects of the invention, a reference frequency database is provided in which the reference frequency for each field for each document registered in the document database is registered as the evaluation data. In the search processing means, when the reference frequency is selected as one of the options of the evaluation scale, the document extracted by the search condition is collated with a reference frequency database to use the reference frequency as the evaluation scale. It is characterized by arranging.

According to a sixth aspect of the invention, in the second or third aspect of the invention, the utility value in each field of each document registered in the document database is registered as the document importance serving as the evaluation data. The document importance database and the reference frequency database in which the reference frequency for each field for each document registered in the document database is registered as the evaluation data are provided. When the evaluation point obtained by weighting and adding the appearance frequency of the keyword, the document importance and the reference frequency is selected as one of the options of the evaluation scale, the document extracted by the search condition is stored in the document importance database and the reference frequency database. The evaluation points are obtained by matching and matching the appearance frequency of the keyword in each extracted document. The evaluation points, characterized in that arranging the documents using the rating scale.

According to the invention of claim 7, in the invention of claim 5 or 6, when the document registered in the document database is referenced in the search processing means, the document is registered in the reference frequency database. It is characterized in that a larger value is added to the reference frequency as the reference frequency is higher.

According to an eighth aspect of the present invention, in the fifth or sixth aspect of the present invention, the search processing means votes when the document registered in the document database is referred to, which corresponds to the utility value to the user. It is characterized in that a value is entered and the sum of voting values is used as the reference frequency.

According to a ninth aspect of the present invention, in the fourth or sixth aspect of the invention, the retrieval processing means votes when the document registered in the document database is referred to, which corresponds to the utility value to the user. A value is input, and the document importance of the document registered in the document importance database is corrected so as to increase as the voting value increases.

The invention of claim 10 is a method for searching for a target document from a document database in which a plurality of documents used within a specific plurality of fields are stored. After specifying the fields of the range, documents that match the search conditions of the set search sentence are extracted from the document database, and further, in each field of each document obtained from the evaluation data classified by field for each document. Let the user select an evaluation scale from multiple types of different methods of obtaining, and then calculate the type of evaluation scale selected for the specified field for each extracted document, and sort the extracted documents in order of magnitude. It is characterized in that they are arranged and output as search results.

According to the invention of claim 11, in the invention of claim 10, after the user inputs the primary search sentence which is a natural sentence including the keyword, the keyword is extracted from the primary search sentence and becomes the next keyword. Generates a secondary search sentence using the related terms by collating the keywords extracted from the primary search sentences with the related term database in which the related terms for the vocabulary are registered in the above-mentioned fields, and uses them from the primary search sentence and the secondary search sentence. It is characterized by allowing a person to select the search sentence.

According to a twelfth aspect of the present invention, in the tenth or eleventh aspect of the invention, the document importance, which is the utility value in each field for each document registered in the document database, and the document database, Using at least one of the reference frequency of each registered document for each field and the appearance frequency of the keyword in each document extracted from the document database according to the search condition as the evaluation scale, the document is identified. Characterized by lining up.

According to the invention of claim 13, in the invention of claim 12, when a document registered in the document database is referred to, a user inputs a voting value corresponding to the utility value, and the total of the vote values is calculated. It is characterized by being used for reference frequency.

According to the invention of claim 14, in the invention of claim 12, when the document registered in the document database is referred to, the user inputs a voting value corresponding to the utility value, and the document importance of the document is important. It is characterized in that the degree is corrected so that the larger the voting value, the larger the degree.

[0026]

BEST MODE FOR CARRYING OUT THE INVENTION In the present embodiment, a server 1 having a document database DB1 storing documents and a terminal 2 operated by a user to search for documents stored in the document database DB1 are in a local area. Although an example in which the server 1 and the terminal 2 are connected via a network NT is shown, the technical idea of the present invention is to connect the server 1 and the terminal 2 via a wide area network such as the Internet, or to the server 1 without using the network. It is also applicable when the function of the terminal 2 is realized by one computer device. In the present embodiment, the process from product planning to sales, that is, the case of searching for documents used in each business of product planning, development, manufacturing, and sales is described as an example. Product development, product design, manufacturing technology, quality control, product sales, and trouble recording are assumed, but even for other documents, documents in multiple fields are registered in the document database DB1 and the documents are used for each field. In that case, the technical idea of the present invention can be applied. For example, in addition to fields by business classification, fields by user's department within the company can be used as fields.

Each document stored in the document database DB1 need not be provided with a keyword when the full text search is possible, but when the full text search is impossible such as when the document is image data. Is appropriately added with keywords and explanations. That is, the vocabulary included in the keyword or the description is searched for in this type of document. Further, in the document database DB1, each document is associated with a headline.

As shown in FIG. 1, the server 1 comprises a search statement setting means 11 for setting a search statement for extracting a document stored in the document database DB1, and a search set by the search statement setting means 11 By giving the sentence to the retrieval processing means 12, the document accumulated in the document database DB1 is collated with the retrieval condition designated by the retrieval sentence. The search sentence is given in the form of a natural sentence as described later. The search processing means 12 extracts a vocabulary to be a keyword included in the search sentence, and if there are a plurality of vocabularies to be a keyword, extracts a logical combination of the vocabulary. Here, the vocabulary logical connection relation means a combination of a logical product, a logical sum, and a negation, which can be extracted by semantic analysis of a search sentence. Thus, the search condition can be set by extracting the vocabulary and the logical connection of the vocabulary from the search sentence. The search processing means 12 not only collates the documents registered in the document database DB1 with the search conditions, but also extracts the document importance database DB2 in which each document is associated with the document importance and each document in the past. The document to be extracted is determined by referring to the reference frequency database DB3 in which the created frequencies are associated. However, this processing will be described later.

The information about the document extracted by the search processing means 12 is stored in the search result output means 13 and presented from the search result output means 13 to the terminal 2. On the screen of the display device provided in the terminal 2, the headlines of the documents extracted by the search processing means 12 are displayed in a list, and if a document that matches the user's search intention is specified from the displayed headlines, the list is specified. The document is read from the document database DB1 and transferred to the terminal 2. In this way, since the headline of the document extracted by the search processing means 12 is presented to the terminal 2 and only the document corresponding to the headline selected by the user is transferred to the terminal 2, the number of documents transmitted through the local network NT. It is possible to transfer a document without significantly increasing the traffic even when the data size of the document is large.

By the way, the search sentence setting means 11 carries out the morphological analysis of the primary search sentence and the primary search sentence input means 14 in which the primary search sentence designated by the user is inputted in the form of a natural sentence. And a search sentence analysis unit 1 for setting a keyword (word or compound word) expanded to include a related term related to the extracted keyword.
5, the secondary search sentence generation means 16 for generating a secondary search sentence of a natural sentence using the keyword set in the search sentence analysis means 15, the primary search sentence and the secondary search sentence are presented to the user. And an execution search sentence selection means 17 for selecting a search sentence desired by the user when searching for a document. Also,
The search sentence setting means 11 is referred to when the search sentence analyzing means 15 sets a part-of-speech database DB4 in which a part of speech of a vocabulary is registered and a keyword including a related term is set in the search sentence analyzing means 15. In order to do so, a related term database DB5 in which related terms are registered is provided. The related terms registered in the related term database DB5 are not only synonyms for the keyword vocabulary, but also technical terms or special terms for each field, or a vocabulary having a specific relevance other than the synonyms for the vocabulary. Is also registered as a related term. For example, generally, a document handled in a company has an important vocabulary such as a vocabulary related to troubles even when it is associated with a disadvantage. Therefore, a vocabulary having such a specific relevance is also registered in the related term database DB5 as a related term. To be done.

In the related term database DB5, not only the related terms for the vocabulary but also the degree of importance of each related term for each field (hereinafter referred to as term importance) is registered in association with each related term. For example, "electrical characteristics"
In the field of quality control, the related terms are "leakage, short circuit, overcurrent, current", in the field of product planning, the related terms are "current, voltage, temperature", and in the field of product design. Is expressed as “current, phase, Lissajous figure, temperature drift”, the related terms are classified according to fields as shown in Table 1, and numerical importance is associated with each related term. In this example, “leakage, short-circuit, overcurrent, temperature drift” is a vocabulary that is associated with disadvantages but is an important vocabulary for “electrical characteristics”. Registered in DB5. Although the method of setting the term importance is not described in detail, it is set based on the number of documents in which the vocabulary is used in each field and the appearance frequency of the vocabulary in one document.

[0032]

[Table 1]

The search sentence analysis means 15 performs part-of-speech decomposition (morphological analysis) of the primary search sentence input through the primary search sentence input means 14, and the vocabulary (mainly nouns, verbs, adjectives, adverbs, etc. Adjective verbs are also possible). If the vocabulary extracted by the part-of-speech decomposition has a compound word, the compound word is extracted. For example, when the primary search sentence is “about the electrical characteristics of the product”, the primary search part is collated with the part-of-speech database DB4,
Part-of-speech decomposition is performed in the form of "product / no / electricity / characteristic / to / about" (where / indicates a part-of-speech division).
Since the primary search sentence includes the compound word "electrical characteristics", the primary search sentence finally becomes "product / of / electrical characteristic /
It is converted to the form By converting the primary search sentence in this way, “product” and “electrical characteristic” are adopted as keywords.

When the vocabulary to be adopted as the keyword is determined from the primary search sentence, each keyword is collated with the related term database DB5, and when the related term exists for the keyword extracted by the search statement analyzing means 15, which keyword is found? The user is presented whether the keyword has a related term. Since the related terms are different for each field, it is possible to specify the field, and the search statement analysis unit 15 collates the related term for the specified field. Here, if the user wants to present related terms for a particular keyword,
A list of related terms for the keyword is presented together with the term importance, and the user can select a desired related term while referring to the term importance. The secondary search sentence generation means 16 generates a secondary search sentence using the keyword extracted from the primary search sentence and the related term selected by the user from the related terms stored in the related database DB5.

For example, the keywords extracted from the primary search sentence as described above are "product" and "electrical characteristic", and the related term database DB5 stores the contents of Table 1 as related terms of "electrical characteristic". If it is assumed to be registered and the user specifies "quality control" as the field, related terms "leakage, short circuit, overcurrent, current" are extracted and presented to the user together with the term importance. It If the user uses "leakage" and "overcurrent" as related terms, the keywords are "product", "leakage", and "overcurrent". A secondary search sentence "About product leakage and overcurrent" is generated.

When the primary search text is input and the secondary search text is generated by the above-mentioned operation, the execution search text selecting means 17 presents the primary search text and the secondary search text to the user. At this stage, the user can select the search sentence to be used for the search from the primary search sentence and the secondary search sentence.
However, in this embodiment, only one search sentence can be selected from the primary search sentence and the secondary search sentence. In this way, not only the primary search text but also the secondary search text is presented to the user, and the secondary search text is also used to enable the search, so that the search by the keyword that the user does not notice can be performed. It will be possible.

Both the primary search text and the secondary search text are natural texts, and the search texts are input to the search processing means 12 regardless of which search text is selected by the user. The search processing means 12 extracts the search condition from the search sentence which is a natural sentence as described above. If the secondary search sentence is used as the search sentence in the above example, "about product leakage and overcurrent" is the search sentence, and therefore the search processing means 12
Then, the vocabulary of "product", "leakage" and "overcurrent" is extracted, and the logical sum of the logical product of "product" and "leakage" and the logical product of "product" and "overcurrent" is used as the search condition. Collate with the document database DB1. In other words, if the logical symbol of the logical product is ∧ and the logical symbol of the logical sum is ∨, the product ∧ (leakage ∨ overcurrent)
Document database DB1 for documents that satisfy the search condition
It is extracted from.

By the way, as described above, the search processing means 12 not only collates the document registered in the document database DB1 with the search condition, but also associates the document importance of each field with each document. A document to be extracted is determined by referring to the document importance database DB2 and a reference frequency database DB3 in which the frequency of each document in the past (hereinafter referred to as a reference frequency) is associated. That is, in the search processing means 12, the frequency with which the keyword used as the search condition appears in the document (hereinafter referred to as the appearance frequency), the document importance of each field of each document, and the reference frequency of each field are used as evaluation data. The evaluation scale of each document is obtained by using it, and the headings of the documents satisfying the search condition are arranged in the descending order of evaluation scale and output to the search result output means 13. As the document importance and the reference frequency for each field, the document importance and the reference frequency for the field specified by the search statement setting unit 11 are used. Document importance database DB2 and reference frequency database D
Data examples with B3 are shown in Table 2 and Table 3, respectively. In addition,
The document importance stored in the document importance database DB2 is set by the document registrant or the like.

[0039]

[Table 2]

[0040]

[Table 3]

As the reference frequency, the frequency at which the body of the document in which the headline is extracted as described above is requested may be used, but in the present embodiment, when the document is browsed (the body of the document is requested. And), a value weighted so that the reference frequency stored in the reference frequency database DB3 increases as the reference frequency increases, and this value is added to the current reference frequency. Further, a value obtained by totaling voting values input by the user after browsing each document may be used as the reference frequency. The former reference frequency increases rapidly as the text is requested more often,
The latter reference frequency is changed by the judgment of the user because the user who browses each document votes the voting value regarding the utility value of the document in multiple stages.

As the evaluation scale, in addition to the evaluation point EP obtained by the following equation, appearance frequency, document importance, reference frequency, etc. can be selected. That is, the user can select a desired evaluation scale from a plurality of types of evaluation scales that are obtained differently from the evaluation data. EP = ω1 × appearance frequency + ω2 × document importance degree + ω3 × reference frequency However, ω1, ω2, ω3 are weighting factors, and are appropriately set according to the calculation method of the appearance frequency, the document importance degree, and the reference frequency. For example, as shown in Table 4, when the appearance frequency is a percentage of the number of appearance times of the keyword with respect to the number of vocabularies in the document, the document importance is a numerical value of 10 levels, and the reference frequency is the reference frequency, ω1 = 50, ω2 = 1, ω3 = 0.05, etc. can be set. Alternatively, when the appearance frequency and the reference frequency are normalized so as to be a numerical value in a range similar to the document importance, ω1 = 1.0, ω2 = 0.8,
It is possible to set ω3 = 1.2 or the like.

[0043]

[Table 4]

As described above, the documents are arranged in the descending order of the evaluation points EP in consideration of the order of the document importance in the field designated by the search sentence setting means 11 and the order of the reference frequency in the field. This means that the documents are arranged in descending order of frequency and frequency of reference. It is also possible to use a value obtained by changing the weighting factor of the evaluation point in the above expression as the evaluation scale. Using the appearance frequency, the document importance, and the reference frequency independently as an evaluation scale, or using any two of them in combination as an evaluation scale is equivalent to setting any weighting coefficient to zero.

The processing procedure in the search processing means 12 is shown in FIG.
Shown in. When the search is started by the search processing means 12, the search condition is collated with the document database DB1 and the search is executed (S1). Search results that match the search conditions are temporarily stored in the memory (S2). Further, the search processing means 12 acquires the designated field from the search sentence setting means 11 (S3), extracts the document importance level corresponding to each document for the acquired field from the document importance level database DB2 (S4), and also refers to it. Refer to frequency Frequency database D
Extract from B3 (S5). The documents stored in the memory are rearranged using the document importance and reference frequency thus obtained, and the result is output to the search result output means 13 (S6).

The operation of this embodiment will be described below by showing a specific work procedure. It is assumed that the screens shown in FIGS. 3 to 10 are displayed on the display device of the terminal 2 connected to the server 1. Before starting the document search, the screen shown in FIG. 3 is first displayed on the terminal 2. On this screen, a field F1 for prompting the input of a primary search sentence, a field F2 for displaying a related term for a keyword extracted from the primary search sentence input in the field F1,
A field F3 for designating a field in which a document is searched is provided. In the vicinity of the field F1, "execute search",
Buttons B1 to B3 for "word expansion" and "reset" are provided. The "execute search" button B1 is used to instruct execution of a document search using only the primary search sentence, the "expand phrase" button B2 is used to specify generation of the secondary search sentence, and the "reset" button B3 is used. Is used when erasing the primary search sentence written in the field F1 and instructing the input of a new primary search sentence.

Below the field F1 presented on the screen of the terminal 2 by the primary search sentence input means 14, there is provided a field F2 for displaying the related terms of the keyword, and above and below the right end of the field F2. A button B4 labeled "Search with secondary search text" is provided. Further, below the left end of the field F2, a field F3 for designating a field is provided. Further, a button B5 is provided adjacent to the field F3. When button B5 is operated (usually, it means clicking with a pointing device such as a mouse),
A pop-up menu corresponding to the field F3 is presented, and when the cursor is placed on the option shown in the pop-up menu and selected (generally, when a pointing device such as a mouse is used, the cursor is placed and clicked. In the case of operation, it means that the return key is pressed after aligning the cursor with the cursor key), and the content of the field F3 is confirmed. The option corresponding to the field F3 is a field related to the document stored in the document database DB1, and in the present embodiment, as described above,
Product planning, product development, product design, manufacturing technology, quality control,
Each field of product sales and trouble recording can be selected. A button B6 described as "logout" is provided in the lower right part of the screen on which the fields F1 and F2 are displayed, and this button B6 is operated when ending the document search processing.

Now, as shown in FIG. 4, when the primary search section "about the electrical characteristics of the product" is entered in the field F1, the morpheme database DB4 is referred to in the search statement analysis means 15 as described above. Analysis is performed and “product” and “electrical characteristics” are extracted as keywords. Here, when the "expand phrase" button B2 is operated, the search statement analysis means 15 collates the keyword with the related term database DB5, and the related term database DB5.
If there is a keyword for which a related term is registered in 5, the keyword is displayed in the “expansion target phrase” field of the field F2, and the “phrase search” button B13 is displayed in the “expansion execution” field of the field F2. As described above, the fact that the "electrical characteristic" is indicated in the field F2 indicates that the related term is registered in the "electrical characteristic".

When the user wants to know a related term regarding the vocabulary of "electrical characteristics", "word expansion button"
By operating B2, as shown in FIG. 5, field F3
A field F5 displaying a list of related terms in the vocabulary “electrical characteristics” for the field selected in is displayed on the screen of the terminal 2. Here, above the field F5, a field F4 is provided which indicates to which vocabulary the related term shown in the field F5 is related. In the illustrated example, the related term of “electrical characteristics” in the field of “product design” is shown in the field F5 (the contents of Table 1 are assumed). Further, the term importance is written in parallel with each related term. A "return" button B7 is provided at the lower right of the field F5, and a "return" button B7 is provided.
It is possible to return to the state of the previous screen by the operation of.

On the screen in which the related terms are shown in the field F5 as shown in FIG. 5, the user can refer to the term importance and select the related terms used for the keyword. That is, as shown in FIG. 6, when a related term to be used as a keyword is selected (generally, the mouse cursor is clicked in the vicinity of the related term), each related term is highlighted (the shaded area in the figure is reversed). Area displayed). The illustrated example shows a state in which "leakage" and "overcurrent" are selected from the related terms. This is because "electric current" is an important vocabulary in "electrical characteristics" in fields such as product design, but it is not so important in the field of "quality control" and rather "leakage" and "overcurrent" are important. Become. Therefore, a user who searches for a document in the field of "quality control" selects "leakage" and "overcurrent".
When the "return" button B7 is operated in a state where "leakage" and "overcurrent" are highlighted, "leakage, overcurrent" is displayed in the "extended phrase" field in the field F2, as shown in FIG.
Is displayed. That is, the primary search sentence is displayed in the field F1 on the screen of the terminal 2, and the user selects the related term database DB2 among the keywords extracted from the primary search sentence.
The related term selected from is displayed in the field F2.
If it is necessary to correct the extended phrase, "extended phrase"
It is also possible to select a field and modify it to another vocabulary.

In this state, "Search by secondary search text" button B
When 4 is operated, the secondary search sentence generating means 16 automatically generates a secondary search sentence which is a natural sentence. When the secondary search text is generated, the execution search text selection means 17 displays the secondary search text in the field F6 in the newly opened window W1, as shown in FIG. Here, since "leakage" and "overcurrent" are selected as related terms, a secondary search sentence "about product leakage and overcurrent" is generated. In this window W1, the user is asked whether or not to search the document using the generated secondary search text, and the user asks for the “Yes” button B in the window W1.
8 or the "No" button B9 is operated. Window W by operating "Yes" button B8
1 is closed and the search using the secondary search sentence is automatically executed. If the "No" button B9 is operated, the window W1 is closed and the previous screen is displayed. On the previous screen, the primary search sentence is displayed in the field F1, so the "execute search" button B1
You can search by the primary search sentence by operating.

When one of the primary search sentence and the secondary search sentence is selected and the execution of the search is instructed, the search processing means 12 searches the document database DB1 and refers to the document importance database DB2 and the reference frequency database DB3. Then, documents satisfying the conditions are extracted, and the above-mentioned evaluation points EP are used to arrange the headings of the extracted documents in descending order of evaluation points EP. That is, as shown in FIG. 9, a search sentence (here, a secondary search sentence) in which a search is performed is shown in the field F1, and a field F7 displayed below the field F1 shows the type of evaluation scale and The calculation method is shown. In the illustrated example, since the evaluation point EP is used as the evaluation scale, the arithmetic expression of the evaluation point EP is shown. A field F8 displayed below the field F7 displays a list of document locations (file names), evaluation points EP, and document headings. Field F
A "display" button B10 corresponding to each document is provided on the document 8, and the body of the document with the headline is displayed on the screen by operating the "display" button B10.

In the field F11 provided in the lower left of the field F8 in the screen shown in FIG. 9, the user is allowed to select the evaluation scale for arranging the extracted documents. That is, the button B14 is provided in parallel in the field F11, and the button B
When you operate 14, a pop-up menu will be displayed. In this pop-up menu, in addition to “evaluation point”, “appearance frequency”, “document importance”, “reference frequency”
Different types of evaluation scales are available. The documents extracted by the search processing means 12 are sorted according to the evaluation scale selected from the pop-up menu,
The user can sort the document headings in the order of various evaluation scales, and can select various methods for finding documents that match the search intention. The field F1
The rating scale displayed in 1 is also shown in the field F7.

After the document matching the retrieval intention is extracted from the document database DB1 by the above-mentioned operation, as shown in FIG. 10, a field F9 indicating a field is displayed.
And a field F10 indicating the extracted document is displayed. This screen has a column for entering the user's vote value for the extracted document, and votes the document's utility value in three stages: "Very useful,""Useful," and "Not very useful." It is supposed to do. Radio button B12 is attached to each stage item,
The voting value is input by selecting one of the radio buttons B12. The voting value entered on this screen is the reference frequency database DB as described above.
It is used to calculate the reference frequency of 3. That is, the added value of the reference frequency is increased with respect to the vote value corresponding to "very useful", and the added value of the reference frequency is decreased with respect to the vote value corresponding to "not very useful". Also,
The voting value is also used as a correction value for the document importance stored in the document importance database DB2. That is,
The document importance database DB2 has an area for storing a correction value set based on a vote value in association with each document for each field, and the correction value set based on the vote value is used in subsequent searches. Will be done. This correction value is set to increase the document importance for the vote value corresponding to "very useful", and to reduce the document importance for the vote value corresponding to "not very useful". Is set to do.

FIG. 11 shows the entire processing procedure of this embodiment described above. That is, when trying to search for a document from the document database DB1, the user first inputs a primary search sentence (S1). The input primary search sentence is part-of-speech decomposed (S2), and when the related term for the keyword extracted from the primary search sentence is registered in the related term database DB5 (S3), it is designated by the user to use the related term. Then, a secondary search sentence is automatically generated (S4). The generated secondary search sentence is presented to the user and the user is made to select whether or not to use the secondary search expression (S
5). If there is no related term here, it will be searched using the primary search sentence. When the search formula is determined, the search is executed (S6), and when a plurality of documents are extracted as the search result (S7), the documents are sorted in descending order of evaluation scale (S).
8). If there is only one extracted document, it is output as it is. When the user is requested to browse the body of the document thus extracted, it is considered that the document has been referred to (S9), and the reference count is updated (S10). If it is not referenced, the process ends as it is.

[0056]

According to the invention of claim 1, a document database in which a plurality of documents used within a range of a plurality of specific fields are stored, and a search statement for allowing a user to set a search statement and specify a field of the search range. And a search result output unit for outputting a search result by the search processing unit, a search processing unit for extracting a document that matches the search condition of the search text set by the search text setting unit from the document database. The search processing means has a function of obtaining an evaluation scale in each field of each document based on the evaluation data classified by field for each document registered in the document database, and a desired one from a plurality of types of evaluation scales different in how to obtain the evaluation scale. A function to let the user select an evaluation scale, and to calculate an evaluation scale of the type selected by the user for the fields specified by the search sentence setting means for each extracted document And has a function of arranging the evaluation scales in order of magnitude and passing them as search results to the search result output means, and outputting the search results in order of magnitude of evaluation scales by using the evaluation scales in which the utility value of each document is evaluated in each field. Therefore, it is possible to preferentially present documents that are likely to match the user's search intention, and the target document can be extracted without considering the noise that is unnecessary for the user. Will be more likely. Moreover, since the user selects a desired evaluation scale from multiple types of evaluation scales, the order of the documents can be changed by using different evaluation scales. The chances of reaching a document can be increased.

According to a second aspect of the present invention, in the first aspect of the present invention, the search sentence setting means allows the user to input a primary search sentence that is a natural sentence including a keyword, and a primary search sentence. A search sentence analyzing means for extracting a keyword from the related sentence, a related term database in which the related terms for the vocabulary to be the keyword are registered for each of the fields, and the keyword extracted by the search sentence analyzing means are collated with the related term database to be searched by the search sentence analyzing means. A secondary search sentence generation means having a function of generating a secondary search sentence using a related term for the extracted keyword, and a search sentence given to the search processing means from the primary search sentence and the secondary search sentence to the user. It consists of the execution search sentence selection means for selecting, and since the related terms for expanding the keyword are classified by field, the keyword is expanded. While reducing the search omission, it is possible to reduce the possibility that contain unwanted noise by limiting the field.

According to the invention of claim 3, in the invention of claim 2, in the related term database, in addition to the synonyms of the vocabulary as a keyword, a vocabulary having a specific relevance in each field is registered as a related term. At the same time, each related term is associated with a term importance that is a measure of the utility value in each field, and the secondary search sentence generation means shows the term importance to the user together with the related term, and the related term This is to select the desired related term, and since the term importance for each field is set for the related term, it is possible to give the user a measure of the effectiveness of the vocabulary to be expanded as a related term by the term importance. Since the user can select the expansion range of the keyword, it is possible to reduce the number of extracted documents that are noises that do not match the search intention. That is, there is a high possibility that a document that matches the user's search intention can be extracted.

According to the invention of claim 4, in the inventions of claims 1 to 3, the utility value in each field of each document registered in the document database is registered as the document importance which is the evaluation data. A document importance database is provided, and when the document importance is selected as one of the evaluation scale options in the search processing means, the document extracted by the search condition is collated with the document importance database to thereby determine the document importance. Documents are arranged using the degree as the evaluation scale, and the utility value of the document in each field is set as the document importance. Therefore, if the registrant sets the document importance, The utility value of each document in each field can be arranged according to a person's intention, and a document having a higher utility value is more likely to be presented in a higher rank.

According to a fifth aspect of the present invention, in the first to third aspects of the present invention, a reference frequency database is provided in which the reference frequency for each field for each document registered in the document database is registered as the evaluation data. In the search processing means, when the reference frequency is selected as one of the options of the evaluation scale, the document extracted by the search condition is collated with a reference frequency database to use the reference frequency as the evaluation scale. Documents that are frequently referred to, that is, documents that have been used frequently are presented in a higher order.

According to the invention of claim 6, in the invention of claim 2 or claim 3, the utility value in each field of each document registered in the document database is registered as the document importance serving as the evaluation data. The document importance database and the reference frequency database in which the reference frequency for each field for each document registered in the document database is registered as the evaluation data are provided. When the evaluation point obtained by weighting and adding the appearance frequency of the keyword, the document importance and the reference frequency is selected as one of the options of the evaluation scale, the document extracted by the search condition is stored in the document importance database and the reference frequency database. The evaluation points are obtained by matching and matching the appearance frequency of the keyword in each extracted document. Documents are arranged by using the evaluation points as the evaluation scale, and the documents are presented in a rank in which the appearance frequency, the document importance, and the reference frequency are comprehensively considered, and the target document has a high rank. Will be more likely to be presented at.

According to the invention of claim 7, in the invention of claim 5 or 6, when the document registered in the document database is referenced in the search processing means, the document is registered in the reference frequency database. The larger the reference frequency is, the larger value is added to the reference frequency, and the ranking can be performed with the reference frequency being emphasized.

According to an eighth aspect of the present invention, in the fifth or sixth aspect of the invention, the search processing means votes corresponding to the utility value to the user when the document registered in the document database is referenced. A value is entered, and the total of voting values is used as the reference frequency. Since the order of presentation is determined by the judgment of the utility value by the user, the higher the value of the document, the higher the value of the document. Become.

According to a ninth aspect of the present invention, in the fourth or sixth aspect of the present invention, the search processing means votes when the document registered in the document database is referred to, which corresponds to the utility value to the user. A value is entered, and the document importance of the document registered in the document importance database is corrected to be larger as the voting value is larger. The document importance is corrected by the judgment of the utility value by the user. Therefore, it becomes easy to extract the document corresponding to the value judgment of the user.

The invention of claim 10 is a method for retrieving a target document from a document database in which a plurality of documents used within a specific plurality of fields are stored. After specifying the fields of the range, documents that match the search conditions of the set search sentence are extracted from the document database, and further, in each field of each document obtained from the evaluation data classified by field for each document. Let the user select an evaluation scale from multiple types of different methods of obtaining, and then calculate the type of evaluation scale selected for the specified field for each extracted document, and sort the extracted documents in order of magnitude. The feature is that they are output side by side as a search result, and the search results are arranged in order of magnitude of the evaluation scale using an evaluation scale that evaluates the utility value of each document by field. Therefore, it is possible to preferentially present documents that are likely to match the user's search intention, and the target document can be extracted without considering the noise that is unnecessary for the user. Will be more likely. Moreover, since the user selects a desired evaluation scale from multiple types of evaluation scales, the order of the documents can be changed by using different evaluation scales. The chances of reaching a document can be increased.

According to the invention of claim 11, in the invention of claim 10, after the user inputs the primary search sentence which is a natural sentence including the keyword, the keyword is extracted from the primary search sentence and becomes the next keyword. Generates a secondary search sentence using the related terms by collating the keywords extracted from the primary search sentences with the related term database in which the related terms for the vocabulary are registered in the above-mentioned fields, and uses them from the primary search sentence and the secondary search sentence. It is characterized by allowing the user to select the above-mentioned search sentence, and the related terms for expanding the keyword are classified by field. Therefore, it is unnecessary to restrict the field while expanding the keyword to reduce the omission of search. The possibility that noise is included can be reduced.

According to a twelfth aspect of the invention, in the tenth or eleventh aspect of the invention, the document importance, which is the utility value in each field for each document registered in the document database, and the document database are stored. Using at least one of the reference frequency of each registered document for each field and the appearance frequency of the keyword in each document extracted from the document database according to the search condition as the evaluation scale, the document is identified. The feature is that they are arranged, and when the document importance is used as the evaluation scale, the documents can be extracted in the order according to the intention of the person who set the document importance, and when the reference frequency is used as the evaluation scale, the usage history of the document is improved. The documents can be extracted in the order of conformity, and if the appearance frequency is used as the evaluation scale, the documents can be extracted in the order of conformity to the objective evaluation scale of the document itself. Therefore, if an evaluation scale that integrates these is used, it is highly possible that the target document can be arranged and presented in a higher order.

According to the invention of claim 13, in the invention of claim 12, when the document registered in the document database is referred to, the user inputs a voting value corresponding to the utility value, and the total of the vote values is calculated. It is characterized in that it is used for reference frequency. Since the order of presentation is determined by the judgment of the utility value by the user, the documents recognized as having higher utility value are presented higher.

According to a fourteenth aspect of the present invention, in the twelfth aspect of the present invention, when the document registered in the document database is referred to, the user inputs a voting value corresponding to the use value, and the document importance of the document is important. The feature is that it is corrected so that the larger the voting value is, the larger the value becomes, and the document importance is corrected by the judgment of the utility value by the user, so it becomes easier to extract the document corresponding to the user's value judgment. .

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is an operation explanatory diagram of a search processing means used in the above.

FIG. 3 is an operation explanatory diagram of the above.

FIG. 4 is an operation explanatory diagram of the above.

FIG. 5 is an operation explanatory diagram of the above.

FIG. 6 is an operation explanatory diagram of the above.

FIG. 7 is an operation explanatory diagram of the above.

FIG. 8 is an operation explanatory diagram of the above.

FIG. 9 is an explanatory diagram of an operation of the above.

FIG. 10 is an operation explanatory diagram of the above.

FIG. 11 is an operation explanatory view showing the overall processing procedure of the above.

[Explanation of symbols]

1 server 2 terminals 11 Search sentence setting means 12 Search processing means 13 Search result output means 14 Primary search text input means 15 Search sentence analysis means 16 Secondary Search Statement Generation Means DB1 document database DB2 document importance database DB3 Reference frequency database DB4 part-of-speech database DB5 related term database

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

[Procedure amendment]

[Submission date] November 28, 2001 (2001.11.
28)

[Procedure Amendment 1]

[Document name to be corrected] Drawing

[Correction target item name] Figure 9

[Correction method] Change

[Correction content]

[Figure 9]

Claims (14)

[Claims] 1. A document database in which a plurality of documents used within a range of a plurality of specific fields are stored, a search statement setting unit for allowing a user to set a search statement and a field of a search range, and a search statement setting. The document processing apparatus includes search processing means for extracting a document that matches the search condition of the search sentence set by the means from the document database, and search result output means for outputting the search result by the search processing means. The function to obtain the evaluation scale in each field of each document based on the evaluation data classified by field for each document registered in, and the user can select the desired evaluation scale from multiple types of evaluation scales with different methods. Function, and for each extracted document, obtain the evaluation scale of the type selected by the user for the field specified by the search sentence setting means, and A document retrieval system having a function of arranging in a small order and delivering the retrieval result to a retrieval result output means. 2. The search sentence setting means, a primary search sentence inputting means for allowing a user to input a primary search sentence which is a natural sentence including a keyword, a search sentence analyzing means for extracting a keyword from the primary search sentence, and a keyword. A secondary search using a related term database in which related terms for the vocabulary to be registered are registered in the above-mentioned fields, and a keyword extracted by the search statement analysis unit is collated with the related term database and related terms for the keyword extracted by the search statement analysis unit are used. A secondary search sentence generating means having a function of generating a sentence; and an execution search sentence selecting means for causing a user to select a search sentence to be given to the search processing means from a primary search sentence and a secondary search sentence. The document retrieval system according to claim 1, wherein the document retrieval system is a document retrieval system. 3. In the related term database, in addition to the synonyms of a vocabulary that is a keyword, a vocabulary having a specific relevance in each field is registered as a related term, and each related term has a utility value in each field. Is associated with the degree of importance of terms, and the secondary search sentence generation means shows the degree of importance of terms to the user together with the related terms and selects the related terms desired by the user from the related terms. The document search system according to claim 2. 4. A document importance database in which the utility value of each document registered in the document database in each field is registered as the document importance serving as the evaluation data, and the retrieval processing means has the document importance. Is selected as one of the options of the evaluation scale, the documents extracted by the search conditions are collated with the document importance database to arrange the documents by using the document importance as the evaluation scale. The document search system according to any one of claims 1 to 3. 5. A reference frequency database in which a reference frequency for each field for each document registered in the document database is registered as the evaluation data, wherein the search processing means has a reference frequency as an option of the evaluation scale. When selected as one, the documents extracted by the search condition are collated with a reference frequency database to arrange the documents by using the reference frequency as the evaluation measure.
The document search system according to claim 3. 6. A document importance database in which the utility value in each field of each document registered in the document database is registered as the document importance serving as the evaluation data, and each document registered in the document database And a reference frequency database in which the reference frequency of each field is registered as the evaluation data, and the search processing means performs weighted addition of the appearance frequency of the keyword in each extracted document, the document importance, and the reference frequency. When the evaluation point is selected as one of the options of the evaluation scale, the document extracted by the search condition is collated with the document importance database and the reference frequency database, and the appearance frequency of the keyword in each extracted document is obtained. The evaluation points are obtained by using the evaluation points and the documents are arranged by using the evaluation points as the evaluation scale. The document search system according to claim 2 or 3, which is used as a signature. 7. The retrieval processing means, when a document registered in the document database is referred to, adds a larger value to the reference frequency as the reference frequency registered in the reference frequency database increases. The document search system according to claim 5 or 6, which is characterized. 8. The search processing means causes a user to input a voting value corresponding to a utility value when a document registered in the document database is referenced, and uses the total of the vote values as a reference frequency. The document search system according to claim 5 or 6, which is characterized. 9. The search processing means causes a user to input a voting value corresponding to a utility value when a document registered in the document database is referenced, and the document registered in the document importance database. 7. The document retrieval system according to claim 4 or 6, wherein the document importance of is corrected so as to increase as the voting value increases. 10. A method for retrieving a target document from a document database in which a plurality of documents used within a range of a specific plurality of fields are stored, wherein a user sets a search sentence and designates a field of the search range. After that, the documents that match the search conditions of the set search text are extracted from the document database, and the evaluation scale in each field is calculated from the evaluation data classified by field for each document. The user is asked to select from a plurality of different types, and then the evaluation scale of the selected type is determined for the specified field for each extracted document, and the extracted documents are arranged in order of magnitude of the evaluation scale and output as a search result. A document retrieval method characterized by: 11. A relation in which a user inputs a primary search sentence that is a natural sentence including a keyword, extracts a keyword from the primary search sentence, and then a related term for the vocabulary to be the keyword is registered for each of the fields. A feature is that a keyword extracted from a primary search sentence is collated with a term database to generate a secondary search sentence using related terms, and a user is allowed to select the search sentence from the primary search sentence and the secondary search sentence. The document search method according to claim 10. 12. A document importance, which is a utility value in each field for each document registered in the document database, and a reference frequency for each field for each document registered in the document database, 11. The document is arranged by using at least one of the appearance frequency of a keyword in each document extracted from the document database according to a search condition as the evaluation scale.
Document search method described in 1. 13. The method according to claim 12, wherein when a document registered in the document database is referred to, a user inputs a voting value corresponding to a utility value, and the total of the vote values is used as a reference frequency. Document search method described. 14. When a document registered in the document database is referred to, the user is made to input a voting value corresponding to the utility value, and the document importance of the document is corrected so as to increase as the voting value increases. 13. The document search method according to claim 12, wherein: