JP7106021B2 - Document retrieval device, program and new attribute value addition method - Google Patents

Description

The present disclosure relates to a document search device, program, and new attribute value addition method.

As a method of searching registered documents to find a document desired by a user, a full-text search for finding a document containing a keyword entered by the user and an attribute value prepared in advance by the user selecting the attribute Attribute searching, which finds documents associated with a value, is known.

For example, in Patent Document 1, among meta information, which is attribute information included in a document created by a user, meta information with a high appearance frequency is used as a search condition candidate, and an attribute search is performed using that search condition. A document retrieval device is disclosed that enables

JP 2008-234214 A

However, the conventional technology only selects meta information with a high appearance frequency as a search condition candidate, and does not reflect the contents of a search performed by a user.

Accordingly, an object of one or more aspects of the present disclosure is to enable a new attribute value to be added according to the content of a search performed by a user.

A document search device according to an aspect of the present disclosure refers to attribute search index information that associates each of a plurality of documents stored in a document storage unit with an attribute value of each of the plurality of documents, a search unit that finds one or more documents by searching a storage unit; a search date and time that is the date and time when the search is performed by the search unit; and one or more documents found by the search unit. a search history storage unit for storing, as a search history, search events indicated by the search history; generating a new attribute value by analyzing the search history; identifying a document associated with the new attribute value as a related document; and a new attribute value adding unit for adding the related document to the attribute search index information.

A program according to an aspect of the present disclosure causes a computer to refer to attribute search index information that associates each of a plurality of documents stored in a document storage unit with an attribute value of each of the plurality of documents, a search unit that finds one or more documents by searching a document storage unit, a search date that is the date and time when the search is performed by the search unit, and one or more documents that are found by the search unit; a search history storage unit for storing search events shown as a search history; and analyzing the search history to generate a new attribute value, identify a document associated with the new attribute value as a related document, and identify the document associated with the new attribute value as a related document. It is characterized by functioning as a new attribute value addition unit that adds the value and the related document to the attribute search index information.

A new attribute value adding method according to an aspect of the present disclosure is such that a search unit creates attribute search index information that associates each of a plurality of documents stored in a document storage unit with an attribute value of each of the plurality of documents. One or more documents are found by referencing and performing a search of the document store, and a search history store stores the search date and time, which is the date and time when the search was performed, and the documents found in the search. A search event indicating one or a plurality of documents is stored as a search history, and a new attribute value addition unit analyzes the search history to generate a new attribute value and associate a document to be associated with the new attribute value. The document is specified as a document, and the new attribute value adding unit adds the new attribute value and the related document to the attribute search index information.

According to one or more aspects of the present disclosure, attribute values can be newly added according to the content of a search performed by a user.

1 is a block diagram schematically showing the configuration of a document retrieval device according to Embodiments 1-3; FIG. It is a schematic diagram showing an example of a search screen image. (A) and (B) are tables for explaining attributes and attribute values. FIG. 4 is a schematic diagram showing a first example of attribute search index information; It is a schematic diagram showing an example of a search history. 4 is a block diagram schematically showing the configuration of a search event pair extraction unit according to Embodiment 1; FIG. FIG. 4 is a schematic diagram for explaining the relationship between two search events included in a temporally continuous search event pair; (A)-(D) are schematic diagrams showing example distributions of documents associated with candidate attribute values. FIG. 4 is a schematic diagram showing a first example of updated attribute search index information; (A) and (B) are block diagrams showing hardware configuration examples. 4 is a flowchart showing search processing in the document search device; FIG. 10 is a flow chart showing processing in which a new attribute value addition unit extracts a new attribute value and adds it to attribute search index information; FIG. 4 is a flow chart showing a process of extracting a target search event pair by a search event pair extraction unit according to Embodiment 1. FIG. FIG. 11 is a schematic diagram showing a second example of updated attribute search index information; It is a schematic diagram showing a modification of the search screen image. FIG. 12 is a block diagram schematically showing the configuration of a search event pair extraction unit according to Embodiment 2; FIG. FIG. 13 is a flow chart showing a process of extracting a target search event pair by a search event pair extraction unit according to Embodiment 2; FIG. FIG. 12 is a block diagram schematically showing the configuration of a document retrieval device according to Embodiment 4;

Embodiment 1.
FIG. 1 is a block diagram schematically showing the configuration of a document retrieval device 100 according to Embodiment 1. As shown in FIG.
The document search device 100 includes an input unit 101, a search input processing unit 102, a document database (hereinafter referred to as a document DB) 103, a search unit 110, a search result display processing unit 120, a display unit 121, and search history processing. It includes a section 122 , a search history storage section 123 , and a new attribute value addition section 130 .

The input unit 101 receives an instruction input from the user.
For example, the input unit 101 receives at least one of an input of a search keyword, which is a keyword for searching documents, and an input of a search attribute value, which is an attribute value for searching, from the user. The attribute value is a value representing the classification of the bibliographic information or description content of the document.

The search input processing unit 102 causes the display unit 121 to display a search screen image, and receives an instruction input from the user via the input unit 101 on the search screen image.

FIG. 2 is a schematic diagram showing an example of a search screen image.
As shown in FIG. 2, the search screen image 140 includes a first attribute value selection area 140a, a second attribute value selection area 140b, a search keyword input area 140c, and a search execution instruction input area 140d. , and a search result display area 140e.

The first attribute value selection area 140a is an area where the user selects a first attribute value included in the first attribute. For example, the user uses the input unit 101 to select the first attribute value from the pull-down menu of the first attribute value selection area 140a. The first attribute and first attribute value will be described later.
The second attribute value selection area 140b is an area where the user selects a second attribute value included in the second attribute. For example, the user uses the input unit 101 to select the second attribute value from the pull-down menu of the second attribute value selection area 140b. The second attribute and second attribute value will be described later.

The first attribute value selected in the first attribute value selection area 140a is also referred to as the first search attribute value, and the second attribute value selected in the second attribute value selection area 140b is referred to as the second attribute value. It is also called a search attribute value of 2.

The search keyword input area 140c is an area where the user inputs a search keyword. For example, the user uses the input unit 101 to enter a search keyword in the search keyword input area 140c.
The search execution instruction input area 140 d is an area where the user inputs an instruction to execute a document search to the document search apparatus 100 . For example, the user selects an attribute value and/or inputs a search keyword in at least one of the first attribute value selection area 140a, the second attribute value selection area 140b, and the search keyword input area 140c. By doing one of the above and pressing the search execution instruction input area 140d using the input unit 101, the document search apparatus 100 can be caused to execute a document search.

The search result display area 140e displays the search result of the document executed by the document search apparatus 100. FIG. The search result display area 140e will be described later.

FIGS. 3A and 3B are tables for explaining attributes and attribute values in this embodiment.
FIG. 3A is a table showing first attributes and first attribute values that are attribute values of the first attributes.
As shown in FIG. 3A, in this embodiment, the first attribute is "model", and the first attribute value is to identify the first attribute "model". is a "model name" as model identification information. Here, the "model name" is "product A", "product B" or "product C".

FIG. 3B is a table showing second attributes and second attribute values that are attribute values of the second attributes.
As shown in FIG. 3B, in this embodiment, the second attribute is "query category" and the second attribute value identifies the second attribute "query category". "Inquiry segment name" as query segment identification information for Here, the "inquiry category name" is "about installation", "function and usage", "maintenance" or "in case of trouble".

The search input processing unit 102 provides the search unit 110 with at least one of the first attribute value, the second attribute value, and the search keyword input by the user.
Further, the search input processing unit 102 provides the search history processing unit 122 with at least one of the first attribute value, the second attribute value, and the search keyword input by the user as a search condition.

Returning to FIG. 1, the document DB 103 is a document storage unit that stores a plurality of documents to be searched by the document search device 100. FIG.

The search unit 110 receives at least one of a search attribute value and a search keyword from the search input processing unit 102 and searches the document DB 103 to retrieve one or more documents related to the received content. find out from
Note that in the present embodiment, the search unit 110 can perform a search using at least one of the first attribute value, the second attribute value, and the search keyword. Embodiments are not limited to such examples. The search unit 110 only needs to be able to search the document DB 103 using at least one attribute value.

The search unit 110 includes a full-text search index storage unit 111 , a full-text search unit 112 , an attribute search index storage unit 113 , an attribute search unit 114 and a search result generation unit 115 .
The full-text search index storage unit 111 stores full-text search index information that associates search keywords with documents containing the search keywords. If each time a search keyword is entered, all documents stored in the document DB 103 are accessed and searched using the search keyword, the search time increases as the number of documents increases. For this reason, based on the results of pre-searching all the documents stored in the document DB 103 in advance using an assumed search keyword, for example, full-text search traction information that associates a search keyword with a document containing that search keyword is prepared. By doing so, full-text search can be performed efficiently. The search keywords stored in the full-text search index information may be in word units, character units, or combinations thereof. Note that the full-text search index storage unit 111 for storing full-text search index information is not necessarily required, and may be omitted.

The full-text search unit 112 searches the document DB 103 using the search keyword given from the search input processing unit 102 to find documents containing the search keyword.
For example, the full-text search unit 112 refers to the full-text search index information stored in the full-text search index storage unit 111 to identify documents containing the search keyword given from the search input processing unit 102 . Then, the full-text search unit 112 identifies the text of the part containing the search keyword in the identified document. The full-text search unit 112 provides the identified document and the identified text to the search result generation unit 115 .

The attribute search index storage unit 113 stores attribute search index information that associates attribute values with documents.
FIG. 4 is a schematic diagram showing a first example of attribute search index information according to the first embodiment.
As shown in FIG. 4, the attribute search index information 141 has a document column 141a and an attribute value column 141b.
The document column 141a stores document IDs, which are document identification information for identifying documents.
The attribute value column 141b stores at least one of the first attribute value and the second attribute value associated with the document identified by the document ID in the same row.

Returning to FIG. 1, the attribute search unit 114 uses at least one of the first attribute value and the second attribute value given from the search input processing unit 102 to search the document DB 103 to obtain one or more find the document of
For example, the attribute search unit 114 refers to the attribute search index information stored in the attribute search index storage unit 113 so that at least the first attribute value and the second attribute value given from the search input processing unit 102 are Identify the documents associated with either. The attribute search unit 114 then provides the identified document to the search result generation unit 115 .

The search result generation unit 115 generates search results using information provided from at least one of the full-text search unit 112 and the attribute search unit 114 .
For example, when the search is performed only by the full-text search unit 112, in other words, when only the keyword search is performed, the search result generation unit 115 selects the specified document given from the full-text search unit 112. Associated attribute values are specified from the attribute search index information stored in the attribute search index storage unit 113 . Then, the search result generation unit 115 gives the specified attribute value and the specified document and specified text given from the full-text search unit 112 to the search result display processing unit 120 as search results.
Further, when a search is performed only by the attribute search unit 114, in other words, when only an attribute search is performed, the search result generation unit 115 selects the specified document given from the attribute search unit 114. It is given to the search result display processing unit 120 as a search result.
Furthermore, when the search is performed by the full-text search unit 112 and the attribute search unit 114, in other words, when the keyword search and the attribute search are used together, the search result generation unit 115 performs the above-described , both the specified attribute value, the specified document and the specified text when only the keyword search is performed, and the specified document when only the attribute search is performed, the search result It is given to the display processing unit 120 . By combining full-text search and attribute search in this way, the user can freely specify search keywords, but full-text search, which may be somewhat lacking in accuracy, and accurate attribute values by predetermined attribute values. It is possible to perform a search that makes use of the advantages of both attribute searches that can be searched.

Search result display processing unit 120 generates a search result partial image representing the search result based on the search result provided from search result generation unit 115 . The search result display processing unit 120 displays the generated search result partial image as part of the search screen image displayed on the display unit 121 .

For example, the search result display processing unit 120 displays the search result partial image in the search result display area 140e in the search screen image 140 shown in FIG.
The search result display area 140e includes a number column 140e-1, a document reference button column 140e-2, a text column 140e-3, a model column 140e-4, and an inquiry category column 140e-5.

The number column 140 e - 1 displays document numbers for identifying documents included in the search results given from the search result generator 115 . The document numbers are sequentially numbered from the top of the search result display area 140e.

The document reference button row 140e-2 displays document reference buttons, which are display instruction input areas for the user to input an instruction to display the corresponding document on the display section 121 via the input section 101. FIG.
When a document reference button displayed in the document reference button row 140e-2 is pressed via the input unit 101, the search result display processing unit 120 reads the corresponding document from the document DB 103 and displays it on the display unit 121. Displaying it allows the user to browse or refer to it.

The text column 140e-3 displays the text included in the search results given from the search result generator 115. FIG. It should be noted that text is not included in the search results for documents found only by attribute search, so in such a case, the text column 140e-3 is left blank.

The model column 140e-4 displays the model name, which is the first attribute value included in the search results given from the search result generator 115. FIG.
The query category column 140e-5 displays the query category name, which is the second attribute value included in the search results given from the search result generator 115. FIG.

In addition, the search result display processing unit 120 includes the search date and time when the search is performed by the search unit 110, the document of the search result which is the document included in the search result given from the search result generation unit 115, Reference documents, which are documents referred to by the user, are given to the search history processing unit 122 as search result display information. A reference document is also referred to as a browsed document that has been browsed by a user.

The display unit 121 displays various screen images. For example, the display unit 121 displays a search screen image.

The search history processing unit 122 uses the search condition given from the search input processing unit 102 and the search result display information given from the search result display processing unit 120 as a search event to search stored in the search history storage unit 123. Store in history.

The search history storage unit 123 stores search histories.
FIG. 5 is a schematic diagram showing an example of search history.
As shown in FIG. 5, the search history 142 records the date and time of search, search conditions, documents of search results, and reference documents for each search event. In the following description, a document as a search result is also referred to as a document indicated by a search event.

The new attribute value addition unit 130 shown in FIG. 1 analyzes the search history to generate a new attribute value, identify a document to be associated with the new attribute value as a related document, and add the new attribute value and the related document. to the attribute search index information.
For example, the new attribute value adding unit 130 analyzes the search history to extract characteristic keywords that match the user's search intention from documents stored in the document DB 103, and uses the extracted characteristic keywords to Generate new attribute values.

The new attribute value addition unit 130 includes a search event pair extraction unit 131 , an attribute value selection unit 132 , a characteristic keyword extraction unit 133 and a new attribute value generation unit 134 .
The search event pair extraction unit 131 analyzes the search history stored in the search history storage unit 123 and extracts a target search event pair, which is two search events that satisfy predetermined event extraction conditions.

FIG. 6 is a block diagram schematically showing the configuration of the search event pair extraction unit 131. As shown in FIG.
The search event pair extraction unit 131 includes a search event pair selection unit 131a, a search date/time continuity determination unit 131b, and a pre-search and post-search document set comparison unit 131c.

The search event pair selection unit 131a selects two temporally consecutive search events based on the search dates and times of the search events stored in the search history, and selects a search event pair based on the selected two search events. Generate. For example, when search event A, search event B, and search event C are stored in the search history in this order in terms of search date and time, the search event pair selection unit 131a consists of search event A and search event B. A search event pair and a search event pair consisting of search event B and search event C are generated.

The search date/time continuity determination unit 131b excludes search event pairs whose search date/time difference is greater than or equal to a predetermined threshold from the search event pairs generated by the search event pair selection unit 131a. Identify short search event pairs that are pairs of search events that are close to each other. The predetermined threshold is, for example, 10 minutes.

The pre- and post-search document collection comparison unit 131c compares the search result corresponding to the temporally later search event from each of the short-time search event pairs specified by the search date/time continuity determination unit 131b to the temporally earlier search event. , are specified as target search event pairs to be extracted by the search event pair extracting unit 131 .

FIG. 7 is a schematic diagram for explaining the relationship between two search events included in a temporally continuous search event pair.
Here, the first search result shown in FIG. 7 is the document of the search result corresponding to the temporally earlier search event in the search event pair. On the other hand, the second search result is the document of the search result corresponding to the later search event in the search event pair.

Reference G1 shown in FIG. 7 indicates documents included in the first search result but not included in the second search result.
Reference G2 indicates a document included in the first search result and also included in the second search result.
Reference G3 indicates documents included in the second search result but not included in the first search result.
Reference G4 indicates a document that is not included in either the first search result or the second search result among the plurality of documents stored in the document DB 103 .

Here, the pre-search and post-search document set comparison unit 131c sets the first condition represented by the following formula (1), the second condition represented by the following formula (2), and the following formula (3). A short-time search event pair that satisfies the third condition is specified as a target search event pair.
G2N>Th1 (1)
G2N÷(G1N+G2N)<Th2 (2)
G3N÷(G2N+G3N)<Th3 (3)
However, G1N is the number of G1 documents, G2N is the number of G2 documents, and G3N is the number of G3 documents.
Also, Th1 is a predetermined first threshold, and is 30, for example. Th2 is a predetermined second threshold, for example, 0.6. Th3 is a predetermined third threshold, for example, 0.1.

By satisfying the first condition, it is estimated that the first search result and the second search result are related.
By satisfying the second condition and the third condition, it is presumed that the search condition that yields the second search result is a more detailed search condition than the search condition that yields the first search result.
By satisfying these conditions, it can be inferred that the user has narrowed down through these two searches.

As described above, the event extraction condition for the search event pair extraction unit 131 to extract the target search event pair is a predetermined value (here, Then, the first event extraction condition that the condition is less than the threshold time), and the temporally later search event of the two search events included in the search event pair is and a second event extraction condition on the condition that the document indicated by the previous search event is narrowed down.
The second event extraction condition is that the first number, which is the number of documents duplicated in both of the two search events included in the search event pair, exceeds a predetermined first threshold. a predetermined second A value obtained by subtracting the first number from the number of documents indicated by the temporally later search event, of the two search events that are less than the threshold and included in the search event pair, is added to the search event pair. A value obtained by dividing by the number of documents indicated by the temporally later search event of the two included search events is less than a predetermined third threshold.

As described above, the search event pair extraction unit 131 shown in FIG. It can be extracted as a target search event pair.

The attribute value selection unit 132 selects an attribute value that satisfies a predetermined selection condition from the attribute values associated with the document indicated by the target search event pair extracted by the search event pair extraction unit 131 as a new attribute. Select as the target attribute value for which you want to generate a value.

Again, the processing in the attribute value selection unit 132 will be explained using FIG.
The attribute value selection unit 132 identifies each of the attribute values associated with the documents of the search results of the two search target events included in each of the target search event pairs as candidate attribute values. Then, the attribute value selection unit 132 identifies the distribution of documents associated with candidate attribute values in each of the documents G1 to G4 shown in FIG.
Then, the attribute value selection unit 132 selects, as the target attribute value, the candidate attribute value whose specified distribution is shown in FIG. 8A, for example.

FIGS. 8(A)-(D) are schematic diagrams showing example distributions of documents associated with candidate attribute values.
In FIGS. 8A to 8D, each bar graph with symbols G1 to G4 indicates the number of documents with each symbol G1 to G4. The number of documents associated with the value.

FIG. 8A shows a distribution in which many documents associated with candidate attribute values are included in documents with codes G1 and G2, but not many documents are included in documents with codes G3 and G4. .
Candidate attribute values exhibiting such a distribution are relevant to both user-refined search terms, but are not sufficiently differentiated in both of these search terms. , it may be possible to add attribute values corresponding to the user's search criteria.

FIG. 8(B) shows a distribution in which many documents associated with candidate attribute values are included in documents with code G1, but not so many in documents with codes G2 to G4.
Candidate attribute values exhibiting such a distribution are highly likely to be attribute values unrelated to the user's intention of narrowing down the second search.

FIG. 8(C) shows a distribution in which many documents associated with candidate attribute values are included in documents with code G2, but not so many in documents with codes G1, G3, and G4. .
Candidate attribute values showing such a distribution are irrelevant to the first search, and are highly likely to match the user's second search intent for narrowing down. By using such candidate attribute values, the user can perform a search sufficiently, so there is no need to extract new attribute values.

FIG. 8(D) shows a distribution in which documents associated with candidate attribute values are rarely included in documents G1 to G4.
Candidate attribute values exhibiting such a distribution are highly likely to be attribute values unrelated to user searches.

The attribute value selection unit 132 selects, as the target attribute value, the candidate attribute value whose specified distribution is as shown in FIG. A candidate attribute value that satisfies the condition 4, the fifth condition represented by the following formula (5), and the sixth condition represented by the following formula (6) is selected as the target attribute value.
(R1+R2)÷(R3+R4)>Th4 (4)
R1÷G1N>Th5 and R2÷G2N>Th5 (5)
R1/R2>Th6 or R2/R1>Th6 (6)
However, R1 is the number of documents associated with the candidate attribute value among the documents of code G1, and R2 is the number of documents associated with the candidate attribute value of the documents of code G2.
Th4 is a predetermined fourth threshold value, for example, 10. Th5 is a predetermined fifth threshold, eg, 0.8. Th6 is a predetermined sixth threshold, for example, 0.9.

By satisfying the fourth condition, it can be detected that the documents associated with the candidate attribute values are heavily biased toward the documents with the codes G1 and G2.
By satisfying the fifth condition, it can be detected that the documents associated with the candidate attribute value occupy most of the documents with the codes G1 and G2.
By satisfying the sixth condition, it can be detected that the documents associated with the candidate attribute values are not biased among the documents G1 and G2.

Note that when there are a plurality of candidate attribute values that satisfy the fourth to sixth conditions, the attribute value selection unit 132 selects (R1+R2) as the largest candidate attribute value or R2 as the largest candidate attribute value. One candidate attribute value is selected as the target attribute value by a predetermined method such as value.

As described above, the selection condition is the attribute for which the selection condition is determined in the document indicated by the temporally previous search event of the two search events included in the target search event pair. The number of documents associated with the candidate attribute value, which is the value, is determined from a plurality of documents stored in the document DB 103 in the two search events included in the target search event pair in the temporally previous search event. Among the documents excluding the indicated document, the value obtained by dividing by the number of documents associated with the candidate attribute value is higher than a predetermined fourth threshold, and the two included in the target search event pair Among search events, the number of first documents that are documents indicated by temporally earlier search events excluding documents indicated by temporally later search events and the ratio of the number of second documents that are documents associated with the candidate attribute value is higher than a predetermined fifth threshold, and the two search events included in the target search event pair overlap The ratio of the number of fourth documents, which are documents associated with candidate attribute values among the third documents, to the number of third documents, which are documents shown as higher than the threshold and the ratio of the number of the second documents to the number of the fourth documents or the ratio of the number of the fourth documents to the number of the second documents is a predetermined sixth threshold The condition is that it is greater than

The characteristic keyword extraction unit 133 shown in FIG. 1 extracts characteristic keywords, which are characteristic keywords, from the target search event pair whose target attribute value is selected by the attribute value selection unit 132 .
For example, the feature keyword extraction unit 133 extracts a predetermined keyword from a plurality of keywords included in a document indicated by a later search event in a target search event pair whose target attribute value is selected. Keywords that satisfy the conditions are extracted as characteristic keywords.

Here, the feature keyword extracting unit 133 extracts feature keywords using a known technique of tf (term frequency)-idf (inverse document frequency) weight. The tf-idf weight has a smaller value for a keyword that appears in more documents, and a larger value for a keyword that appears only in a specific document.

Again, description will be made with reference to FIG.
The feature keyword extracting unit 133 extracts all keywords included in the document of code G1 in the target search event pair as first keywords. Next, the feature keyword extraction unit 133 calculates the tf-idf weight for each of the first keywords. Then, the feature keyword extraction unit 133 generates a first weighted keyword list in which the first keywords are arranged in descending order of tf-idf weight. In the first weighted keyword list, a first keyword that appears more frequently in the document of code G1 and less frequently appears in other documents is placed at a higher rank.
Next, the feature keyword extraction unit 133 identifies N1 predetermined first keywords from the top in the first weighted keyword list.
Here, N1 is a predetermined positive integer, such as 20. If the number of keywords included in the list is less than N1, the characteristic keyword extraction unit 133 may extract all keywords.

In addition, the feature keyword extraction unit 133 extracts all keywords included in the document of code G2 in the target search event pair as second keywords. Next, the feature keyword extraction unit 133 calculates the tf-idf weight for each of the second keywords. Then, the characteristic keyword extraction unit 133 generates a second weighted keyword list in which the second keywords are arranged in descending order of tf-idf weight. The second weighted keyword list is a list in which second keywords appearing more often in the document of code G2 and less frequently appearing in other documents are arranged at higher ranks.

Next, the characteristic keyword extraction unit 133 identifies M1 predetermined second keywords from the top in the second weighted keyword list. The characteristic keyword extracting unit 133 identifies, among the identified second keywords, the remaining second keywords excluding the keywords that match the first keywords identified as described above, as remaining second keywords. do. Then, the feature keyword extracting unit 133 extracts a predetermined M2 second keywords from the top of the remaining second keywords as feature keywords.

Here, M1 is a predetermined positive integer, such as 20. M2 is also a predetermined positive integer, eg, 5.
It should be noted that the feature keyword extraction unit 133 may extract all keywords when the number of keywords included in the list is less than M1 or M2.

As described above, the keyword extraction condition is the two search events included in the target search event pair in which the target attribute value is selected. one or more keywords identified from the document indicated by the temporally earlier search event of the two search events included in the target search event pair whose target attribute value was selected from the keywords of The condition is to extract a predetermined number of keywords from the remaining keywords in descending order of importance.

The new attribute value generation unit 134 shown in FIG. 1 uses a predetermined method from the target attribute value selected by the attribute value selection unit 132 and the feature keyword extracted by the feature keyword extraction unit 133. to generate a new attribute value. For example, when the target attribute value is "when in trouble" and the feature keyword is "drums", the new attribute value generating unit 134 generates "when in trouble (drums)" as the new attribute value. In addition, when the target attribute value is "when in trouble" and the characteristic keywords are "drums" and "noise", the new attribute value generation unit 134 sets the new attribute value as "when in trouble (drums, noise )”. In this way, when there are multiple characteristic keywords, the new attribute value generation unit 134 may list the multiple characteristic keywords.
As shown in FIG. 3, when there are multiple attributes, the new attribute value is added to the attribute containing the target attribute value used when generating the new attribute value. .

The new attribute value generator 134 then adds the new attribute value to the attribute search index information stored in the attribute search index storage unit 113 .
Here, the document associated with the new attribute value (also referred to as related document) is the G2 document (see FIG. 7) in the target search event pair for which the target attribute value is selected. In other words, a document whose target attribute value is duplicated in both of the two search events included in the selected target search event pair becomes the document associated with the new attribute value. For example, in the target search event when adding the new attribute value "When in trouble (drum)" in the above example, the document D002 shown in FIG. 4 is included in the document G2 shown in FIG. , document D003 is a document other than G2 shown in FIG. is replaced with "when in trouble (drum)", the attribute value "in case of trouble" of the document D003 is maintained.

Some or all of the search input processing unit 102, the search unit 110, the search result display processing unit 120, the search history processing unit 122, and the new attribute value addition unit 130 described above are shown in FIG. As described above, it can be composed of a memory 10 and a processor 11 such as a CPU (Central Processing Unit) that executes programs stored in the memory 10 . Such a program may be provided through a network, or recorded on a recording medium and provided. That is, such programs may be provided as program products, for example.

Part or all of the search input processing unit 102, the search unit 110, the search result display processing unit 120, the search history processing unit 122, and the new attribute value addition unit 130 are shown in FIG. 10B, for example. As such, the processing circuit 12 can be a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

As described above, the search input processing unit 102, search unit 110, search result display processing unit 120, search history processing unit 122, and new attribute value addition unit 130 can be configured by a processing circuit network.
Note that the input unit 101 can be realized by an input device such as a keyboard and a mouse.
The document DB 103, the full-text search index storage unit 111, the attribute search index storage unit 113, and the search history storage unit 123 can be realized by a storage device such as an HDD (Hard Disc Drive).
The display unit 121 can be realized by a display device such as a display.
In other words, the document retrieval device 100 can be realized by a computer.

FIG. 11 is a flow chart showing retrieval processing in the document retrieval apparatus 100. As shown in FIG.
First, the search input processing unit 102 displays a search screen image on the display unit 121, and inputs a keyword for full-text search, a first attribute value for attribute search, and a second attribute value for attribute search through the input unit 101. receives input of at least one of the attribute values of (S10).

Next, the search unit 110 searches for documents stored in the document DB 103 based on the content input in step S10 (S11).
Then, the search unit 110 generates a search result, which is the result of the executed search (S12).

Next, the search result display processing unit 120 displays the search result in part of the search screen image (S13).

Next, the search history processing unit 122 stores the search condition given from the search input processing unit 102 and the search result display information given from the search result display processing unit 120 as a search event in the search history storage unit 123. stored in the search history (S14).

FIG. 12 is a flow chart showing a process in which the new attribute value adding unit 130 generates a new attribute value and adds it to the attribute search index information.
First, the search event pair extraction unit 131 analyzes the search history stored in the search history storage unit 123 and extracts a target search event pair (S20). The processing here will be described in detail with reference to FIG. 13 .

Next, the attribute value selection unit 132 selects one unselected target search event pair from the target search event pairs identified by the search event pair extraction unit 131 (S21).

Next, the attribute value selection unit 132 selects an attribute that satisfies a predetermined condition among the first attribute value and the second attribute value associated with the document of the search result in the selected target search event pair. It is determined whether or not there is a target attribute value that is a value (S22). If there is a target attribute value (Yes in S22), the process proceeds to step S23, and if there is no target attribute value (No in S22), the process returns to step S21.

In step S22, the feature keyword extraction unit 133 extracts feature keywords from the target search event pair for which the target attribute value detected in step S22 was selected.

Next, the new attribute value generation unit 134 generates a new attribute value using the target attribute value detected in step S22 and the feature keyword extracted in step S23, and uses the new attribute value for attribute search. It is added to the attribute search index information stored in the index storage unit 113 (S24).

Next, the attribute value selection unit 132 determines whether or not all target search event pairs have been selected (S25). If unselected target search event pairs remain (No in S25), the process returns to step S21, and if all target search event pairs have been selected (Yes in S25), the process ends. do.

FIG. 13 is a flow chart showing a process of extracting a target search event pair by the search event pair extraction unit 131 according to the first embodiment.
First, the search event pair selection unit 131a selects two temporally consecutive search events from the search history stored in the search history storage unit 123, and selects a search event pair based on the selected two search events. A pair is generated (S30).

Next, the search event pair selection unit 131a selects one search event pair from the generated search event pairs (S31).

Next, the search date and time continuity determination unit 131b determines whether or not the search dates and times of the two search events included in the search event pair selected by the search event pair selection unit 131a are within a short period of time ( S32). For example, if the search date/time continuity determination unit 131b determines that the search dates/times of the two search events are within a short period of time when the difference between the search dates/times of the two search events is less than a predetermined threshold, good. If the search dates and times of the two search events are within a short period of time (Yes in S32), the process proceeds to step S33. , the process returns to step S31.

In step S33, the pre-search and post-search document set comparison unit 131c selects the short-time search event pair that is the target search event pair determined in step S32 that the search date and time are within a short time, that is, that the search is performed close to each other. In the two search events included in , it is determined whether the temporally later search event narrows down the temporally earlier search event. If the two search events have been narrowed down (Yes in S33), the process proceeds to step S34, and if the two search events have not been narrowed down (No in S33), the process returns to step S31.

In step S34, the search event pair selected in step S31 is extracted as a target search event pair.
Then, the retrieved event pair selection unit 131a determines whether or not all retrieved event pairs have been selected (S35). If there are search event pairs that have not been selected yet (No in S35), the process returns to step S31, and if all search event pairs have been selected (Yes in S35), the process ends.

Note that the order in which the processes in steps S32 and S33 in FIG. 13 are performed may be changed.

As described above, according to the first embodiment, it is possible to newly add an attribute value according to the content of the search performed by the user.

In the first embodiment described above, as described with reference to FIGS. 4 and 9, when adding a new attribute value, the target search event pair used to generate the new attribute value The target attribute value associated with the document (that is, the document of G2 in FIG. 7) that is duplicated in both search results is replaced with a new attribute value. It is not limited to such examples. For example, in addition to the target attribute value, the new attribute value may be associated with documents that are duplicated in the search results of both of the target search event pairs used to generate the new attribute value.

Specifically, in the target search event when adding the new attribute value "When in trouble (drum)" in the above example, the document D002 shown in FIG. , and document D003 is a document other than G2 shown in FIG. 7, for example, a document of G1, as shown in the updated attribute search index information 141#2 in FIG. , "When in trouble (drum)" may be added to the attribute value of the document D002 in addition to "When in trouble".

In such a case, for example, like the search screen image 140#1 shown in FIG. It is desirable to have it as a check box rather than a check box.

Embodiment 2.
As shown in FIG. 1, the document search device 200 according to the second embodiment includes an input unit 101, a search input processing unit 102, a document DB 103, a search unit 110, and a search result display processing unit 120. , a display unit 121 , a search history processing unit 122 , a search history storage unit 123 , and a new attribute value addition unit 230 .
The input unit 101, the search input processing unit 102, the document DB 103, the search unit 110, the search result display processing unit 120, the display unit 121, the search history processing unit 122, and the search history storage unit 123 of the document search device 200 according to the second embodiment are , input unit 101, search input processing unit 102, document DB 103, search unit 110, search result display processing unit 120, display unit 121, search history processing unit 122, and search history storage unit 123 of document search device 100 in Embodiment 1 is similar to

The new attribute value addition unit 230 in the second embodiment also analyzes the search history to generate new attribute values, specify related documents, and add the new attribute values and related documents to the attribute search index information. .

The new attribute value addition unit 230 includes a search event pair extraction unit 231 , an attribute value selection unit 132 , a characteristic keyword extraction unit 133 and a new attribute value generation unit 134 .
The attribute value selection unit 132, the characteristic keyword extraction unit 133, and the new attribute value generation unit 134 of the new attribute value addition unit 230 in Embodiment 2 are equivalent to the attribute value selection unit 132 of the new attribute value addition unit 130 in Embodiment 1, It is the same as the feature keyword extraction unit 133 and the new attribute value generation unit 134 .

The search event pair extraction unit 231 analyzes the search history stored in the search history storage unit 123 and extracts a target search event pair that is two search events that satisfy a predetermined condition.

FIG. 16 is a block diagram schematically showing the configuration of the search event pair extraction unit 231 according to the second embodiment.
The search event pair extraction unit 231 includes a search event pair selection unit 131a, a search date/time continuity determination unit 131b, a pre- and post-search document set comparison unit 131c, and a browse presence/absence determination unit 231d.

The search event pair selection unit 131a, the search date/time continuity determination unit 131b, and the pre-search document set comparison unit 131c of the search event pair extraction unit 231 in the second embodiment are similar to the search event pair extraction unit 231 in the first embodiment. This is the same as the pair selection unit 131a, the search date/time continuity determination unit 131b, and the pre-search and post-search document set comparison unit 131c.
However, the search date/time continuity determination unit 131b provides the short-time search event pair to the browse presence/absence determination unit 231d. In addition, the pre-search and post-search document set comparison unit 131c performs the same processing as in the first embodiment using the browsing short-time search event pair instead of the short-time search event pair.

The browsing presence/absence determining unit 231d identifies, as a short browsing search event pair, a pair including the document browsed by the user in the search result document among the short-time search event pairs given from the search date/time continuity judging unit 131b. . For example, the browse presence/absence determination unit 231d confirms the reference information included in the search history 142, so that the document ID is stored in the reference document in at least one of the two search events included in the short-time search event pair. If so, it is determined that it has been viewed by the user.
The browse presence/absence determination unit 231d provides the specified browse short-time search event pair to the pre-search and post-search document set comparison unit 131c.

FIG. 17 is a flowchart showing a process of extracting a target search event pair by the search event pair extraction unit 231 according to the second embodiment.
Among the steps included in the flowchart shown in FIG. 17, processing similar to steps included in the flowchart shown in FIG. attached.

The processing of steps S30 to S32 in FIG. 17 is the same as the processing of steps S30 to S32 in FIG. However, in step S32, if the search dates and times of the two search events are within a short period of time (Yes in S32), the process proceeds to step S40.

In step S40, the browsing presence/absence determination unit 231d determines whether or not any document in the search results of the short-time search event pair provided from the search date/time continuity determination unit 131b has been browsed by the user. Then, if the user has viewed any document (Yes in S40), the process proceeds to step S34, and if the user has not viewed any document (No in S40), the process proceeds to step S34. Return to S31.

The processing of steps S33 to S35 in FIG. 17 is the same as the processing of steps S33 to S35 in FIG. However, in step S33, the pre-search and post-search document set comparing unit 131c temporally Then, it is determined whether or not the later search event narrows down the earlier search event in terms of time.

As described above, in the second embodiment, the event extraction condition is the third event extraction condition that at least one document indicated by two search events included in a search event pair has been browsed. further includes
Note that the order of performing the processes in steps S32, S40 and S33 in FIG. 17 is not specified as the order shown in FIG. 17, and may be performed in any order.

As described above, according to the second embodiment, the presence or absence of the target attribute value is determined only for the search event in which the user browsed the document. is performed and a new attribute value is added.

Embodiment 3.
As shown in FIG. 1, the document search device 300 according to the third embodiment includes an input unit 101, a search input processing unit 102, a document DB 103, a search unit 110, and a search result display processing unit 120. , a display unit 121 , a search history processing unit 122 , a search history storage unit 123 , and a new attribute value addition unit 330 .
The input unit 101, the search input processing unit 102, the document DB 103, the search unit 110, the search result display processing unit 120, the display unit 121, the search history processing unit 122, and the search history storage unit 123 of the document search device 300 according to the third embodiment are , input unit 101, search input processing unit 102, document DB 103, search unit 110, search result display processing unit 120, display unit 121, search history processing unit 122, and search history storage unit 123 of document search device 100 in Embodiment 1 is similar to

The new attribute value addition unit 330 includes a search event pair extraction unit 131 , an attribute value selection unit 132 , a characteristic keyword extraction unit 333 and a new attribute value generation unit 134 .
The search event pair extraction unit 131, the attribute value selection unit 132, and the new attribute value generation unit 134 of the new attribute value addition unit 330 in Embodiment 3 are the same as the search event pair extraction unit of the new attribute value addition unit 130 in Embodiment 1. 131 , the attribute value selection unit 132 and the new attribute value generation unit 134 .

The characteristic keyword extracting unit 333 according to Embodiment 3 extracts characteristic keywords, which are characteristic keywords, from the target search event pair whose target attribute value is selected by the attribute value selecting unit 132 .
In the third embodiment as well, the characteristic keyword extracting unit 333 extracts characteristic keywords using tf-idf weighting, which is a known technique.

Again, description will be made with reference to FIG.
The feature keyword extracting unit 333 extracts all keywords included in the document of code G1 in the target search event pair as first keywords.
Next, the feature keyword extraction unit 133 calculates the tf-idf weight for each of the first keywords. Then, the feature keyword extraction unit 133 generates a first weighted keyword list in which the first keywords are arranged in descending order of tf-idf weight.
Next, the feature keyword extraction unit 333 identifies N1 predetermined first keywords from the top in the first weighted keyword list.

In addition, the feature keyword extraction unit 333 extracts all keywords included in the document G2 in the target search event pair as second keywords. Next, the feature keyword extraction unit 333 calculates the tf-idf weight for each of the second keywords. Then, the feature keyword extraction unit 333 generates a second weighted keyword list in which the second keywords are arranged in descending order of tf-idf weight.

Next, when the search keyword is included in the search conditions corresponding to the second search result, the characteristic keyword extraction unit 333 extracts the search keyword included in the search condition corresponding to the second search result. Among them, a search keyword that does not match the search keyword included in the search condition corresponding to the first search result is set as an incremental search keyword. If no search keyword is included in the search conditions corresponding to the first search result, all search keywords included in the search conditions corresponding to the second search result are incremental search keywords.

Next, the characteristic keyword extracting unit 333 increases the weight of the second keywords included in the second weighted keyword list that match the incremental search keyword. For example, the characteristic keyword extraction unit 333 increases the keyword weight by a predetermined ratio (for example, two times) and updates the second weighted keyword list to generate an updated weighted keyword list. do.

Next, the feature keyword extraction unit 333 identifies M1 predetermined second keywords from the top in the updated weighted keyword list. The characteristic keyword extracting unit 333 identifies, among the identified second keywords, the remaining second keywords excluding the keywords that match the identified first keywords as remaining second keywords. do. Then, the feature keyword extracting unit 333 extracts a predetermined M2 second keywords from the top of the remaining second keywords as feature keywords.

As described above, when the characteristic keyword extraction unit 333 in Embodiment 3 determines the keyword extraction condition, the time The temporally earlier search event of the two search events included in the target search event pair in which the target attribute value is selected from the multiple keywords included in the document indicated by the later search event In the remaining keywords excluding one or more keywords specified from the document indicated by, the temporally later search of the two search events included in the target search event pair whose target attribute value is selected Among the search keywords used in the event, the importance of keywords matching search keywords that do not match the search keywords used in the previous search event in terms of time is increased by a predetermined method.

As described above, in Embodiment 3, the keywords used for narrowing down the search are more likely to be selected as feature keywords from among the keywords used for the full-text search.

Note that the method of extracting feature keywords in the feature keyword extraction unit 333 is not limited to the above example.
For example, the characteristic keyword extraction unit 333 may extract characteristic keywords as follows.

Again, description will be made with reference to FIG.
The feature keyword extracting unit 333 extracts all keywords included in the document of code G1 in the target search event pair as first keywords. Next, the feature keyword extraction unit 133 calculates the tf-idf weight for each of the first keywords. Then, the feature keyword extraction unit 133 generates a first weighted keyword list in which the first keywords are arranged in descending order of tf-idf weight.
Next, the feature keyword extraction unit 333 identifies N1 predetermined first keywords from the top in the first weighted keyword list.

Next, the feature keyword extracting unit 333 extracts all keywords included in the document of code G2 in the target search event pair as second keywords. Next, the feature keyword extraction unit 333 calculates the tf-idf weight for each of the second keywords. Then, the characteristic keyword extraction unit 133 generates a second weighted keyword list in which the second keywords are arranged in descending order of tf-idf weight.

Next, the feature keyword extraction unit 333 identifies M1 predetermined second keywords from the top in the second weighted keyword list. The characteristic keyword extracting unit 333 selects the remaining second keywords from the identified second keywords, excluding the keywords that match the first keywords identified as described above, as remaining second keywords. Identify.

Then, if there is a search keyword included in the search condition corresponding to the second search result among the remaining second keywords, the characteristic keyword extraction unit 333 extracts such second search result. are identified as feature keywords.
On the other hand, if there is no remaining second keyword that matches the search keyword included in the search condition corresponding to the second search result, the characteristic keyword extraction unit 333 extracts the remaining second keyword. Among them, M2 predetermined second keywords are extracted as feature keywords from the top.

As described above, in this modified example, the keyword extraction condition is the document indicated by the temporally later search event of the two search events included in the target search event pair in which the target attribute value is selected. Among the two search events included in the target search event pair whose target attribute value is selected from the included keywords, one or more specified from the document indicated by the temporally earlier search event match the search keyword used in the temporally later search event of the two search events included in the target search event pair whose target attribute value is selected for the remaining keywords excluding the If there is a matching keyword, extract the matching keyword, and if there is no matching keyword, extract a predetermined number of keywords in descending order of importance from the remaining keywords. This is the condition.

Embodiment 4.
FIG. 18 is a block diagram schematically showing the configuration of a document retrieval device 400 according to Embodiment 4. As shown in FIG.
The document search device 400 includes an input unit 101, a search input processing unit 402, a document DB 103, a search unit 410, a search result display processing unit 120, a display unit 121, a search history processing unit 422, and a search history storage. 423 and a new attribute value addition unit 430 .
The input unit 101, the document DB 103, the search result display processing unit 120, and the display unit 121 of the document search device 400 according to the fourth embodiment are the same as the input unit 101, the document DB 103, and the search result of the document search device 400 according to the first embodiment. It is similar to the display processing unit 120 and the display unit 121 .

The search input processing unit 402 causes the display unit 121 to display a user authentication screen image (not shown) for performing user authentication before displaying the search screen image on the display unit 121, thereby performing user authentication to the user. Accepts an input as to whether or not
When the user performs user authentication, the search input processing unit 402 receives input of user authentication information, which is information required for user authentication, from the user via the input unit 101 . The user authentication information includes, for example, a user ID, which is user identification information for identifying the user, and may include the user's password.
Here, the processing in the document retrieval apparatus 400 when the user does not perform user authentication is the same as that of the document retrieval apparatus 100 according to the first embodiment. Processing will be mainly described.

The search input processing unit 402 causes the display unit 121 to display the search screen image and displays the search screen image in the same manner as in the first embodiment regardless of whether the user is authenticated or not. In the image, an instruction input is received from the user via the input unit 101 .

The search input processing unit 402 provides the search unit 410 with at least one of the first attribute value, the second attribute value, and the search keyword input by the user. When the user has been authenticated, the search input processing unit 402 also gives the user ID to the search unit 410 .
Further, the search input processing unit 402 provides the search history processing unit 422 with at least one of the first attribute value, the second attribute value, and the search keyword input by the user as a search condition. When the user is authenticated, the search input processing section 402 also gives the user ID to the search history processing section 422 .

The search unit 410 receives at least one of the first attribute value, the second attribute value, and the search keyword from the search input processing unit 402, and searches the document DB 103 to retrieve documents related to the received content. is found from the document DB 103 . Here, the search unit 410 also acquires the user ID from the search input processing unit 402 when the user has been authenticated.

The search unit 410 includes a full-text search index storage unit 111, a full-text search unit 112, an attribute search index storage unit 113, an attribute search unit 414, a search result generation unit 115, and an individual attribute search index storage unit 416. .
The full-text search index storage unit 111, the full-text search unit 112, the attribute search index storage unit 113, and the search result generation unit 115 of the search unit 410 according to the fourth embodiment are the same as the full-text search index storage unit 111 of the search unit 110 according to the first embodiment. , the full text search unit 112 , the attribute search index storage unit 113 and the search result generation unit 115 .

The individual attribute search index storage unit 416 stores initial attribute search index information, which is default attribute search index information, and individual attribute search index information, which is attribute search index information for each user ID. The initial attribute search index information is assumed to be the same information as the attribute search index information stored in the attribute search index storage unit 113 before the new attribute value addition unit 430 adds the new attribute value.

When the attribute search unit 414 receives at least one of the first attribute value and the second attribute value without the user ID from the search input processing unit 402, as in the first embodiment, , the attribute search is executed using the attribute search index information stored in the attribute search index storage unit 113 .

On the other hand, when the attribute search unit 414 receives at least one of the first attribute value and the second attribute value together with the user ID from the search input processing unit 402, the individual attribute search index storage unit At 416, it is checked whether individual attribute search index information corresponding to the received user ID is stored.
When such individual attribute search index information is stored, the attribute search unit 414 uses the individual attribute search index information to execute an attribute search.
If such individual attribute search index information is not stored, the attribute search unit 414 duplicates the initial attribute search index information and associates it with the received user ID to obtain the individual attribute search index for that user ID. Information is newly generated, and an attribute search is executed using the individual attribute search index information.

Then, the attribute search unit 414 provides the specified document to the search result generation unit 115 regardless of whether user authentication has been performed.
In addition, when the user is authenticated, the full-text search unit 112 also gives the user ID to the search result generation unit 115, and the search result generation unit 115 uses the individual attribute search information corresponding to the user ID to associate You just need to get the attribute value

The search history processing unit 422 uses the search condition given from the search input processing unit 102 and the search result display information given from the search result display processing unit 120 as a search event to search stored in the search history storage unit 423. Store in history. Here, when the search history processing unit 422 also receives the user ID from the search input processing unit 402, the search history processing unit 422 includes the user ID in the search event and stores it in the search history.

The search history storage unit 423 stores search histories. In the search history according to Embodiment 4, the user ID is also included in the search event.

If user authentication has not been performed, the new attribute value addition unit 430 generates a new attribute value by analyzing all search events included in the search history, as in the first embodiment. Also, identify related documents and add the new attribute values and related documents to the attribute search index information.

On the other hand, when the user is authenticated, the new attribute value adding unit 430 analyzes the search event associated with the user ID stored in the search history, and generates a new attribute value. , specifies the related document, and adds the new attribute value and the related document to the individual attribute search index information corresponding to the user ID.

The new attribute value addition unit 430 includes a search event pair extraction unit 431 , an attribute value selection unit 132 , a characteristic keyword extraction unit 133 and a new attribute value generation unit 434 .
The attribute value selection unit 132 and the characteristic keyword extraction unit 133 of the new attribute value addition unit 430 according to the fourth embodiment are the same as the new attribute value addition unit 130 according to the first embodiment.

The processing in the search event pair extraction unit 431 when user authentication has not been performed is the same as in the first embodiment.
When the user is authenticated, the search event pair extraction unit 431 analyzes the search event including the authenticated user ID among the search events stored in the search history, and satisfies a predetermined condition. Extract the target search event pair that satisfies. Specific processing in the search event pair extraction unit 431 is the same as the processing in the first embodiment, except that search events including authenticated user IDs are used.

If user authentication has not been performed, the new attribute value generation unit 434 generates a new An attribute value is generated, and the new attribute value is added to the attribute search index information stored in the attribute search index storage unit 113. FIG.

On the other hand, when the user is authenticated, the new attribute value generation unit 434 uses the target attribute value selected by the attribute value selection unit 132 and the feature keyword extracted by the feature keyword extraction unit 133 to , generates a new attribute value and adds the new attribute value to the individual attribute search index information associated with the user-authenticated user ID stored in the individual attribute search index storage unit 416 .

As described above, according to the fourth embodiment, a new attribute value can be added individually for each user ID.

In the first to fourth embodiments described above, document DB 103, full text search index storage unit 111, attribute search index storage unit 113, and individual attribute search index storage unit 416 are provided in document search devices 100 to 400. However, Embodiments 1 to 4 are not limited to such examples. For example, when the document retrieval apparatuses 100 to 400 are connected to a network (not shown) and are capable of communicating with the network via a communication unit (not shown), the document DB 103, full-text search, A part or all of the index storage unit 111, the attribute search index storage unit 113 and the individual attribute search index storage unit 416 may be provided by another device such as a server connected to the network.
Also, the input unit 101 and the display unit 121 may be provided in another device connected to the network. In this case, the search input processing units 102 and 402 send the search screen image to other devices via the network to acquire information necessary for the search, and the search result display processing unit 120 displays the search result partial image. , to other devices to display the search results.

Although the third embodiment described above combines features of the third embodiment with the first embodiment, the third embodiment is not limited to such an example. For example, the features of the third embodiment may be combined with the second embodiment.
Also, in the fourth embodiment, the features of the fourth embodiment may be combined with the second or third embodiment.

100,200,300,400 document search device 101 input unit 102 search input processing unit 103 document DB 110,410 search unit 111 full text search index storage unit 112 full text search unit 113 attribute search index storage unit 114, 414 attribute search unit, 115 search result generation unit, 416 individual attribute search index storage unit, 120 search result display processing unit, 121 display unit, 122 search history processing unit, 123 search history storage unit, 130, 230, 330, 430 new attribute value addition unit 131,231,431 search event pair extraction unit 132 attribute value selection unit 133,333 feature keyword extraction unit 134,434 new attribute value generation unit.

Claims (15)

one or more of a search unit for locating documents;
a search history storage unit that stores, as a search history, a search date and time, which is the date and time when a search was performed by the search unit, and a search event indicating one or more documents found by the search unit;
By analyzing the search history, a new attribute value is generated, a document associated with the new attribute value is specified as a related document, and the new attribute value and the related document are added to the attribute search index information. and an attribute value addition unit. The new attribute value adding unit
From the search history, a search event pair, which is two search events with consecutive search dates and times, is specified, and from the specified search event pairs, a search event pair that satisfies a predetermined event extraction condition is set as a target search event pair. a search event pair extraction unit for extraction;
an attribute value selection unit that selects, as a target attribute value, an attribute value that satisfies a predetermined selection condition from the attribute values associated with the document indicated by the target search event pair;
Among the target search event pairs for which the target attribute value is selected, keywords that satisfy predetermined keyword extraction conditions are selected from a plurality of keywords included in documents indicated by later search events in terms of time as feature keywords. A feature keyword extraction unit that extracts as
generating the new attribute value by a predetermined method from the target attribute value and the characteristic keyword, and in both two search events included in the target search event pair from which the target attribute value is selected; and a new attribute value generating unit that identifies a document shown in duplicate as the related document, and adds the new attribute value and the related document to the attribute search index information. 1. The document retrieval device according to 1. The event extraction condition is
a first event extraction condition on condition that a difference between search dates and times of two search events included in the search event pair is less than a predetermined value;
A second event extraction under the condition that, of the two search events included in the search event pair, the temporally later search event narrows down the document indicated by the temporally earlier search event. 3. The document retrieval device according to claim 2, comprising: a condition; The second event extraction condition is
a first number, which is the number of documents duplicated in both of the two search events included in the search event pair, is greater than a predetermined first threshold;
A value obtained by dividing the first number by the number of documents indicated by a temporally earlier search event out of two search events included in the search event pair is a predetermined second threshold. less than and
The search event pair includes a value obtained by subtracting the first number from the number of documents indicated by the temporally later search event of the two search events included in the search event pair. 4. A value obtained by dividing by the number of documents indicated by the temporally later search event of the two search events is less than a predetermined third threshold. Document retrieval device as described. wherein the event extraction conditions further include a third event extraction condition on condition that at least one document indicated by two search events included in the search event pair has been viewed. 5. The document retrieval device according to claim 3 or 4. The selection condition is
Among the two search events included in the target search event pair, it is associated with the attribute value for which the selection condition is to be determined in the document indicated by the temporally previous search event. The number of documents is the number of documents excluding the document indicated by the temporally previous search event in the two search events included in the target search event pair from the plurality of documents stored in the document storage unit. Among them, a value obtained by dividing by the number of documents associated with the attribute value for which the selection condition is to be determined is higher than a predetermined fourth threshold,
A first document which is a document obtained by excluding the document indicated by the temporally later search event from the document indicated by the temporally earlier search event, out of the two search events included in the target search event pair. The ratio of the number of second documents among the first documents, which are documents associated with the attribute value for which the selection condition is to be determined, to the number of documents is a predetermined number. higher than a threshold of 5,
the number of third documents that are duplicated in two search events included in the target search event pair, among the third documents as targets for determining the selection condition; a percentage of the number of fourth documents that are documents associated with the attribute value that is higher than the fifth threshold, and
A ratio of the number of the second documents to the number of the fourth documents or a ratio of the number of the fourth documents to the number of the second documents is larger than a predetermined sixth threshold 6. The document retrieval device according to any one of claims 2 to 5, wherein the condition is that . The keyword extraction condition is a plurality of keywords included in a document indicated by a temporally later search event out of two search events included in the target search event pair in which the target attribute value is selected. , excludes one or more keywords specified from the document indicated by the temporally previous search event of the two search events included in the target search event pair in which the target attribute value is selected from 7. The document retrieval device according to any one of claims 2 to 6, wherein the condition is that a predetermined number of keywords are extracted from the remaining keywords in descending order of importance. further comprising an input unit that receives input of a search attribute value that is an attribute value for searching,
The search unit is
an attribute search index storage unit that stores the attribute search index information;
8. An attribute search unit that refers to the attribute search index information and finds a document associated with the search attribute value from the document storage unit. The document retrieval device described in . further comprising an input unit that receives input of a search attribute value that is an attribute value for searching,
The search unit is
an attribute search index storage unit that stores the attribute search index information;
7. An attribute search unit that refers to the attribute search index information and finds a document associated with the search attribute value from the document storage unit. The document retrieval device described in . The input unit further receives input of a search keyword that is a keyword for searching,
The search unit is
a full-text search unit for searching the document storage unit for documents containing the search keyword;
a search result generation unit that sets at least one of the documents found by the attribute search unit and the documents found by the full-text search unit as the one or more documents. 10. The document retrieval device according to claim 9. The keyword extraction condition is a plurality of keywords included in a document indicated by a temporally later search event out of two search events included in the target search event pair in which the target attribute value is selected. , excludes one or more keywords specified from the document indicated by the temporally previous search event of the two search events included in the target search event pair in which the target attribute value is selected from A condition that extracts a predetermined number of keywords from the remaining keywords in descending order of importance,
When determining the keyword extraction condition, the characteristic keyword extracting unit determines, in the remaining keywords, the temporal among the search keywords used in the later search event, the importance of keywords that match the search keywords that do not temporally match the search keywords used in the previous search event are increased by a predetermined method. 11. The document retrieval device according to claim 10. The keyword extraction condition is a plurality of keywords included in a document indicated by a temporally later search event out of two search events included in the target search event pair in which the target attribute value is selected. , excludes one or more keywords specified from the document indicated by the temporally previous search event of the two search events included in the target search event pair in which the target attribute value is selected from Among the remaining keywords, among the two search events included in the target search event pair in which the target attribute value is selected, a keyword that matches the search keyword used in the later search event in terms of time. If there is, the condition is to extract the matching keyword, and if there is no matching keyword, the condition is to extract a predetermined number of keywords in descending order of importance among the remaining keywords. 11. The document retrieval device according to claim 10, wherein: The input unit can also accept input of user identification information for identifying a user,
When the user identification information is input, the search unit stores each of the plurality of documents associated with the user identification information and stored in the document storage unit, and searching the document storage unit using individual attribute search index information that associates each attribute value;
when the user identification information is input, the search history storage unit stores the search event in association with the user identification information;
When the user identification information is input, the new attribute value addition unit analyzes a search event associated with the user identification information in the search history and generates the new attribute value. and specifying the related document, and adding the new attribute value and the related document to the individual attribute search index information associated with the user identification information. 1. The document retrieval device according to claim 1. the computer,
one or more of a search unit for locating documents,
a search history storage unit that stores, as a search history, a search date and time, which is the date and time when a search was performed by the search unit, and a search event indicating one or more documents found by the search unit;
By analyzing the search history, a new attribute value is generated, a document associated with the new attribute value is specified as a related document, and the new attribute value and the related document are added to the attribute search index information. A program characterized by functioning as an attribute value adding unit. A search unit refers to attribute search index information that associates each of a plurality of documents stored in a document storage unit with an attribute value of each of the plurality of documents, and executes a search of the document storage unit. to find one or more documents,
A search history storage unit stores, as a search history, a search event indicating a search date and time, which is the date and time when the search was performed, and one or more documents found in the search,
A new attribute value addition unit analyzes the search history to generate a new attribute value and identifies a document to be associated with the new attribute value as a related document,
A new attribute value adding method , wherein the new attribute value adding unit adds the new attribute value and the related document to the attribute search index information.

Images