JPH11184888A - Method for retrieving document and device therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To present a retrieved result obtained by a retrieval processing using a neural network type word data base to a user so that it can be made intuitively understandable. SOLUTION: A retrieving part 16 connects words to form a network, develops a retrieval key based on the strength of connection with a retrieval key designated by a user which is inputted from an inputting part 11 by referring to a neural network type word data base 15 having the direction, kind, and strength of connection for each connection, retrieves a document in a document file 14 by using the developed retrieval key, and obtains a retrieved result. A retrieved result classifying part 17 classifies the retrieved results for each word of the word data base 15 by using the retrieval key. A retrieved result reconstituting part 18 reconstitutes classified retrieved results like a network so that the same relation as that of the network structure of the word data base 15 can be maintained, and arranges it on a map, while a control part 13 presents the map through a display part 12 to a user.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document search apparatus and method for performing a document search by acquiring a word associated with a search key specified by a user as a search key, and more particularly to a search which can be easily understood by a user. The present invention relates to a document search apparatus and method suitable for presenting a result.

[0002]

2. Description of the Related Art In recent years, with the development of OCR (optical character reading device) and the like, documents that have been printed up to now can be held in large quantities as electronic files. As a matter of course, the document file becomes large in size, and a search system for obtaining desired information from the document file becomes necessary.

[0003] Some search systems employ various methods, such as a method of attaching a keyword to a document file in advance, and a full-text search for searching by looking at the contents of the document file. In either case, in order to bridge the gap between the search key entered by the user and the word that appears in the actual document file, a method such as expanding the search key using not only the search key word itself but also synonyms and thesaurus In order to obtain a search result closer to what the user wants, a device has been devised.

[0004] Also, in presenting search results, a device has been devised so that a user can easily find what he or she wants, for example, by arranging the search results as close as possible to those specified by the user.

[0005]

However, in the above-mentioned prior art, a synonym in which words are grouped or a thesaurus having a hierarchical structure is simply used as a kind of grouping. Only a little additional information was given, but no further information was reflected. For this reason, there is room for improvement in obtaining search results closer to what the user wants.

Accordingly, the present inventor has disclosed in Japanese Patent Application No. Hei 9-18851.
In No.6, by using a neural network type word database that connects words to each other in a network type and can specify the type and strength of connection, related words such as complex related words from a search key specified by the user We also proposed a search method to obtain more suitable search results by performing document search using these search keys. However, in the invention (prior application invention) according to the application of Japanese Patent Application No. 9-188516, a visual device that the user can intuitively understand is not sufficiently provided in a method of presenting the search result to the user. .

Accordingly, the present invention has been made to solve the above-mentioned problem, and relates to a search using a neural network type word database which can connect words to each other in a network type and define the strength of connection and the like. It is an object of the present invention to provide a document search apparatus and a method capable of intuitively presenting a search result obtained by a process to a user intuitively.

[0008]

SUMMARY OF THE INVENTION According to the present invention, words are connected to each other in a network, and a user-designated word database is referred to by referring to a neural network type word database having the direction, kind and strength of connection for each connection. The search key is expanded based on the strength of the connection with the search key of, and the search result is obtained by searching the document using the expanded search key.
The acquired search results are presented to the user while maintaining the relationship between the search results using a network-type map that maintains the same relationship as the network structure of the word database.

As described above, a user is not limited to a mere synonym or thesaurus, but is developed by using a neural network type word database that connects words to words in a network type and can specify the strength of connection and the like. Related keywords, such as complicated related words, can also be acquired as search keys from the search key specified by the user, and a more appropriate search result can be obtained by performing document search using this search key.

In addition, by presenting the search results using the neural network type word database in a network-type map that is intuitively understandable, it is possible to present the relevance of the search results to the user in an easy-to-understand manner. Becomes Here, it is assumed that the expanded search key includes the input search key itself (as a search key having the strongest connection with the input search key), but may not include the input search key. In this case, a document is searched using the input search key and the expanded search key.

Further, the present invention provides a method for presenting a search result using a neural network type word database while maintaining the relationship between the search results on a network type map maintaining the same relation as the network structure of the database. In other words, when reconstructing and presenting based on the network structure of the database, the hierarchical relationship on the network structure is regarded as a hierarchy, and the search result by the search key on the same hierarchy as the word specified by the user is obtained. It is also characterized in that it is selected as a presentation target and presented to the user.

As described above, in the neural network type word database, attention is paid to the hierarchical relationship (superordinate / subordinate relationship) having a conceptual hierarchy, and only the relationship at the same level is expressed. It is possible to present the relation of each changed word in an easily understandable manner.

According to the present invention, when retrieving a search result using a neural network type word database based on the network structure of the database and presenting the result, the hierarchical relationship on the network structure is determined. It is also characterized in that it is regarded as a hierarchy, and for each hierarchy, the result of the search key on that hierarchy is presented to the user with a unique modification added to the hierarchy.

As described above, in the neural network type word database, the upper and lower layers having a conceptual hierarchy
By paying attention to the lower-level relations and expressing the relations at the same level separately from the others, the relations for each word serving as the search key can be presented in an easily understandable manner.

Further, according to the present invention, when a search result obtained by using a neural network type word database is reconstructed based on the network structure of the database and presented, a network structure for a user-specified word is used. It is also characterized in that a search result using a search key having a user-specified relationship is selected as a presentation target and presented to the user.

As described above, by selectively presenting search keys having a relationship designated by the user in the neural network type word database, search results desired by the user can be efficiently presented.

[0017]

Embodiments of the present invention will be described below with reference to the drawings. [First Embodiment] FIG. 1 is a block diagram showing a schematic configuration of a document search apparatus according to a first embodiment of the present invention.

In FIG. 1, reference numeral 11 denotes an input unit for inputting a character string, 12 denotes a display unit for displaying a character string input by the input unit 11 and a search result, and 13 denotes a control unit for controlling the entire apparatus. (Main control unit).

Reference numeral 14 denotes a document file to be searched, 1
Reference numeral 5 denotes a word database having search key development and weighting information. Here, the word database 15 is a neural network type database in which words are connected in a network form, and the direction, type and strength of the connection are provided for each connection.

Reference numeral 16 refers to the word database 15 and expands the search key based on the strength of the connection with the search key manually input by the input unit 11, and converts the document from the document file 14 using the expanded search key. It is a search unit for searching.

Reference numeral 17 denotes a search result classifying unit for classifying search results for each word in the word database 15 using a search key;
Reference numeral 8 denotes a search result reconstructing unit that arranges search results classified by the search result classifying unit 17 on a map according to the network structure of the word database 15.

FIG. 2 shows the configuration of the search unit 16. As shown in the figure, a search control unit 161 controls the overall control for the search process. The search unit 16 refers to the word database 15 using the search key input from the input unit 11 and searches for the search key. The search key expansion unit 162 expands a search key based on the strength of connection, and a document search unit 163 searches the document file 14 using the search key expanded by the search key expansion unit 162.

Next, an outline of the operation of the document search apparatus having the configuration shown in FIG. 1 will be described. First, according to the operation of the user, the input unit 1
From 1 is input a word as a search key. This search key is sent to the search unit 16 via the control unit 13 and received by the search control unit 161 in the search unit 16. The search control unit 161 converts the input search key into the search key expanding unit 1
Pass to 62. The search key expanding unit 162 refers to the word database 15 and expands the input search key into the input search key and a plurality of similar search keys.

The word database 15 is a neural network type database in which words are connected in a network form as described above, and the type and strength of the connection are defined. FIG. 3 shows an example of the data structure of the word database 15. The arrows in the figure indicate the direction of connection between the corresponding words, and the type of connection (synonym,
Antonyms, related words, superordinate concepts, subordinate concepts, etc.) and numerical values indicating the strength of the connection. Here, it is shown that the smaller the numerical value, the stronger the connection. Although not shown in the figure, the lower concept has the same strength in the opposite direction as the higher concept.

The search key expansion unit 162 refers to the word database 15 (the search key expansion unit 16
2), the plurality of expanded search keys are stored in the document search unit 16.
Sent to 3. The document search unit 163 searches the document file 15 which is a set of documents to be searched based on the passed search key. The result searched by the document search unit 163 includes information indicating which word in the word database 15 has been matched as a search key, together with information on the search control unit 16.
1 to the search result classifying unit 17. The search result classification unit 17 classifies (groups) search results according to the matched search keys.

The search result reconstructing unit 18 reconstructs the search results classified by the search result classifying unit 17 in accordance with the network structure of the word database 15 and obtains the result of the search processing. The control unit 13 presents the reconstruction result from the search result reconstruction unit 18 to the user through the display unit 12.

Next, the search key expanding section 16 in the search section 16
The details of the process 2 will be described with reference to the flowchart of FIG. First, the search key expanding unit 162 puts the input search key (search word) itself into the search key list (step A1). Next, the search key developing unit 162 searches the word database 15 with the input search word (step A2). If there is no input search word (word matching with) in the word database 15 (step A3), the search key expanding unit 1
At step 62, the addition to the search key list ends.

On the other hand, if the input search word is in the word database 15 (step A3), the search key developing unit 162 sequentially follows the network of the word database 15 (step A4). As is clear from the example of FIG.
Since a word can have multiple connections,
The search key expanding unit 162 traces them all (step A5). However, if you follow a distance, you will be able to follow a word that has little to do with it. If there is no strong connection to some extent, stop following it (step A).
6).

Here, the "degree" of "somewhat"
For example, it is determined by the lower limit of the connection strength. In the word database 15, when the connection is strong, the word database 15 has a small numerical value. Therefore, the "degree" is determined by the maximum value indicating the strength of the connection. Therefore, while following the network of the word database 15, a limit is set such that the total value of the strength of connection from the input search word does not exceed 8, for example. If the condition is satisfied and the connection is strong, the search key list uses the word as a search key and the connection strength (sum of the values)
(Step A7). This is repeated until all necessary networks have been traced. When the input search word (search key) is entered in the search key list in step A1, (minimum value) 0 is set as the value of the strength of association of the input search word with itself. .

A specific example of the processing of the search key expanding section 162 will be described for a case where the word "Belzebub" is input as a search word (search key). First, the input “Belzebub” is registered in the search key list. Further, since the word "Beelzebub" is present in the word database 15 as shown in FIG. 3, the search is performed by following the network.

In the example of the word database 15 shown in FIG.
The superordinate concept of "Beelzebub" is "Devil". This "demon" has a connection strength value of "Beelzebub" of 1
Therefore, this is also added to the search key list. In addition, from the "devil" there is a connection to "angel" and "fallen angel",
In any case, since the sum of the values of the strength of connection from “Beelzebub” does not exceed 8 as in 1 + 5 = 6, 1 + 2 = 3, it is added to the search key list.

On the other hand, there is "Christism" as a superordinate concept from "Angel", but this is not added to the search key list because the total value of the strength of connection exceeds 8 as 1 + 5 + 6 = 12. .

The search key expanding section 162 repeats this to obtain a search key list as shown in FIG. 5 for the input of the search key "Beelzebub".
The document search unit 163 searches the word database 15 according to the search keys on the search key list.

Next, details of the processing of the search result reconstructing unit 18 will be described with reference to the flowchart of FIG. First,
The results searched by the search unit 16 (the document search unit 163 therein) are classified by the search result classifying unit 17 for each search key. The search result reconstructing unit 18 includes the search result classifying unit 1
7, the input unit 1
The search result using the search key input from No. 1 is first arranged on a map (search result map) (step B1).

Next, the search result reconstructing unit 18 checks whether or not processing has been completed for all search results (step B2). If there are unprocessed search results, the search result reconstructing unit 18 sequentially follows the network of the word database 15. (Step B3).
Then, the search result reconstructing unit 18 checks whether or not the searched word exists in the search result (as a search key) (step B4). If the word exists in the search result, the word, that is, the search key is displayed on the map. Is arranged together with the search result on the map so that the relation with the already arranged search key in the above maintains the same relation as the network structure of the word database 15 (step B5).

The search result reconstructing unit 18 repeats the above processing (steps B4 to B7) until all the search results are arranged on the map (step B3). FIG. 7 shows an example of a search result when “Belzebub” is input as a search key (search word). In the example of FIG. 7, there is no document corresponding to “Beelzebub”, but the superordinate concept “Devil”
It can be seen that there were five documents. The search results are classified by the search result classifying unit 17 for each search key. FIG. 8 shows a map display example (search result presentation example) when the classified search results are reconfigured by the search result reconstructing unit 18 according to the network structure of the word database 15 and arranged on a map. . Each word in FIG. 8 has a search result (search result) using each as a search key. For example, “Angel” has a search result as shown in FIG. 9, and the user can refer to it by specifying “Angel” while the map of FIG. 8 is displayed.

When a search key is presented to the user via the display unit 12, the details of the search result can be presented together. FIG. 10 shows an example of the search result presentation.

As described above, when the search results are presented in the same form as the neural network type word database, the relation between the search keys can be easily understood, and the desired search result can be easily found. [Second Embodiment] FIG. 11 is a block diagram showing a schematic configuration of a document search device according to a second embodiment of the present invention.

The document search device shown in FIG. 11 is composed of the input unit 11, the display unit 12, the control unit 13, the document file 1 shown in FIG.
4. Input unit 2 corresponding to word database 15, search unit 16, search result classification unit 17, search result reconstruction unit 18
1, display unit 22, control unit 23, document file 24, word database 25, search unit 26, search result classification unit 27,
It has a search result reconstruction unit 28. The feature of the document search device in FIG. 11 is that, in addition to the above-described components, attention is paid to the upper / lower relationship (upward / downward relationship) of words in the word database 25, and the search result reconstruction unit 28 performs reconstruction. And an upper / lower relation selection unit 29 for selecting a search result by a search key on the same hierarchy as a word specified by the user.

Now, in the document search apparatus having the configuration shown in FIG. 11, the search / rearrangement selector 29 reconfigures the search results classified by the search result classifier 27 by the search result reconstructing unit 28 and arranges them on the map. At this time, only search results using search keys (search words) in the same hierarchy as the user-specified word (center word) in the hierarchical relationship of the corresponding search keys (search words) are selected as the target search results. . Here, the word specified by the user is a word input as a search key from the input unit 21 according to a user operation in an initial state.

The control unit 23 determines that only the reconstructed result in the retrieval result reconstructing unit 28, that is, only the retrieval key (the retrieval result by the retrieval key) in the same hierarchy as the word specified by the user (the central word) is in the network of the word database 25. The map arranged with the same relationship as the structure is presented to the user through the display unit 22. At that time, the control unit 23 also presents a word of a higher concept or a lower concept of each search key together with the search key (search word).

Further, the control unit 23 determines whether a word of a higher concept or a lower concept of the search key (search word) presented corresponding to a certain search key (search word) is operated by the user (for example, When the word is selected and specified from the input unit 21 by a mouse operation), the search result reconstructing unit 28 causes the search result reconstructing unit 28 to perform the reconstructing process again by using the word as a user-specified word (main word). At this time, the upper / lower relation selecting unit 29 selects, from among the search results classified by the search result classifying unit 27,
Only search results using search keys (search words) in the same hierarchy as a new user-specified word (center word) are selected.
Therefore, the high-level concept presented through the display unit 22,
Alternatively, when the word of the lower concept is selected and designated by the user, the display is switched to the presentation of the search result by the search key (search word) in the same hierarchy as the selected word.

Next, details of the processing performed by the search result reconstruction unit 28 and the hierarchical relationship selection unit 29 will be described with reference to the flowchart of FIG. First, results searched by the search unit 26 are classified by the search result classifying unit 27 for each search key. The search result reconstructing unit 28 includes the search result classifying unit 2
7, the search key that is currently focused on, that is, the central search key (in the initial state,
A search result based on a search key input from the input unit 21 according to a user operation is first arranged on a map (search result map) including the search key (step C1).

Next, the search result reconstructing unit 28 checks whether or not processing has been completed for all search results (step C2). If there are unprocessed search results, the search result reconstructing unit 28 sequentially follows the network of the word database 25. (Step C3).
Then, the search result reconstructing unit 28 checks whether or not the searched word is present as a search key in the search result (step C4).

Then, the upper / lower relationship selecting unit 29 determines the word (search key) of the destination followed by the search result reconstructing unit 28 and the search key (word) currently centered on the network structure of the word database 25. Is checked, and the result is notified to the search result reconstruction unit 28 (step C5).

If there is no hierarchical relationship between the two,
That is, if the tracing destination word is on the same hierarchy as the search key (word) that is currently the center (step C
6), the search result reconstructing unit 28 converts the previous word, that is, the search key, so that the relation with the already arranged search key on the map maintains the same relation as the network structure of the word database 25. It is arranged on the map together with the search result (step C7). On the other hand, if there is an upper / lower relationship between the two, that is, if the tracing destination word is not on the same level as the current search key (word), the search result reconstructing unit 28 The process returns to step C2 without arranging on the map.

Retrieval result reconstructing section 28 and hierarchical relation selecting section 2
No. 9 repeats the above processing (steps C3 to C7) until the processing is completed for all search results (step C2).

The control unit 23 determines only the reconstructed result of the search result reconstructing unit 28 performed in cooperation with the upper / lower relationship selecting unit 29, that is, the result of retrieval by the retrieval key in the same hierarchy as the network structure of the word database 25. The map arranged while maintaining the same relationship as is presented to the user through the display unit 22.

FIG. 13 shows a presentation example (map display example) of the retrieval result by the retrieval key on the same hierarchy presented to the user in this manner. FIG. 13A shows an example of presenting a search result when “Beelzebub” is input as a search key (search word) and the input search key is the center. In this example, since “Beelzebub” has no search key other than the upper-lower relation, only this is presented.

In FIG. 13A, when the search key “Belzebub” is presented, it is also presented that the search key “Belzebub” has the word “demon” as a superordinate concept. If the word “demon”, which is a superordinate concept of the search key “Belzebub”, is selected and specified from the input unit 21 by a user operation (for example, a mouse operation), the control unit 23
Selects the term "demon" of the selected superordinate concept, and adds that word "demon" to the new central term (search key)
As a result, the search result reconstructing unit 28 causes the reconstructing process to be performed again. At this time, the upper / lower relation selecting unit 29 selects, from among the search results classified by the search result classifying unit 27, the same hierarchical level as the newly central word, that is, the hierarchical level one level higher than the input search key. Select only search results by a certain search key.

FIG. 13B shows an example of presenting a search result in the case of focusing on the superordinate concept "demon" of "Belzebub".
In this way, by arranging and presenting search keys having no hierarchical relationship in the same hierarchy, it is possible to present the relation between the search keys in an easily understandable manner. [Third Embodiment] FIG. 14 is a block diagram showing a schematic configuration of a document search apparatus according to a third embodiment of the present invention.

The document search device shown in FIG. 14 is composed of the input unit 11, the display unit 12, the control unit 13, the document file 1 shown in FIG.
4. Input unit 3 corresponding to word database 15, search unit 16, search result classifying unit 17, and search result reconstructing unit 18.
1, display unit 32, control unit 33, document file 34, word database 35, search unit 36, search result classification unit 37,
It has a search result reconstruction unit 38. The feature of the document search device in FIG. 14 is that, in addition to the above-described components, the search results classified by the search result classifying unit 37 are focused on the upper / lower relationship (upward / downward relationship) of the words in the word database 35. It has a hierarchical relationship classifying unit 39 for classifying each word in the hierarchical relationship of words on the word database 35.

In the document search apparatus having the configuration shown in FIG. 14, the hierarchical relation classifying unit 39 reconstructs the search results classified by the search result classifying unit 37 and arranges them on a map. At this time, the classification is performed for each layer by focusing on the upper and lower relations of the corresponding search keys (search words). Here, search keys in the same hierarchy are given the same hierarchy identification information unique to the hierarchy, for example, the same hierarchy number. The control unit 33 compares the search results classified by the hierarchical relationship classification unit 39 and reconstructed by the search result reconstructing unit 38 with the search keys in the same hierarchy as the modification (display modification) unique to the hierarchy. For example, the color is changed, the frame line is changed, or the like, and presented to the user through the display unit 32.

Next, details of the processing of the search result reconstructing unit 38 and the hierarchical relation classifying unit 39 will be described with reference to the flowchart of FIG. First, the results searched by the search unit 36 are classified by the search result classifying unit 37 for each search key. The search result reconstructing unit 38 includes the search result classifying unit 3
Among the search results classified by 7, the search result by the search key input from the input unit 31 according to the user operation is first arranged on the map (search result map) including the search key (step D1).

Next, the search result reconstructing unit 38 checks whether or not processing has been completed for all search results (step D2). If there are unprocessed search results, the search result reconstructing unit 38 sequentially follows the network of the word database 35. (Step D3).
Then, the search result reconstructing unit 38 checks whether or not the searched word is included in the search result as a search key (step D4).

Then, the hierarchical relation classifying section 39 sets the input word (search key) followed by the search result reconstructing section 38 and the input search key (word) on the network structure of the word database 25 (input). The hierarchical relationship (based on the retrieved search key) is checked (step D5). Then, the upper / lower relation selecting unit 29 outputs the search key (word) of the above-mentioned destination.
Then, upper and lower hierarchy numbers indicating the upper / lower relationship checked in step D5 are allocated and notified to the search result reconstruction unit 38 (step D6). In this case, if the search key following the input search key is in the same hierarchy without any upper / lower relationship, the search key is “0”, if it is one level higher, “−1”, one level lower. If it is irrelevant, as in the case of "1" and "Lucifer" with respect to "Bersebub" in FIG. 8, upper and lower hierarchical numbers such as the maximum value (or a specific symbol indicating that) are assigned to the search key. .

The search result reconstructing section 28 stores the searched search key on the map so that the relation between the search key and the search key already arranged on the map is the same as the network structure of the word database 35. It is arranged together with the search result (step D7). Here, upper and lower hierarchical numbers are assigned to search keys (and search results using the search keys) arranged on the map.

The search result reconstructing unit 38 and the hierarchical relation classifying unit 3
No. 9 repeats the above processing (steps D3 to D7) until the processing is completed for all search results (step D2).

The control unit 33 classifies the reconstructed result of the search result reconstructing unit 38 performed in cooperation with the hierarchical relationship classifying unit 39, that is, the search key (search result based on the same hierarchy), and In the state where the upper and lower hierarchical numbers unique to are assigned, the search key arranged on the map and the search result by the search key are presented to the user through the display unit 32. At this time, the control unit 33 sets a search key to which the same upper and lower hierarchy numbers are assigned (and a search result using the search key).
Each time the number is given a display modification such as a unique color or a frame line, and presented to the user. That is, the search keys to which the same upper and lower hierarchy numbers are assigned are presented with the same modification.

FIG. 16 shows "Belzebub" as a search key.
9 shows a presentation example of a search result in a case where is input. In this example, as shown in FIG. 15, the search result by the input search key “Belzebub” and the search result by another search key are determined by the vertical relationship with respect to the input search key “Belzebub”. , Are presented with different borders.

As described above, by presenting the search keys on the same hierarchy in the upper-lower relation with the same modification, the relation between the search keys can be presented in an easy-to-understand manner. [Fourth Embodiment] FIG. 17 is a block diagram showing a schematic configuration of a document search apparatus according to a fourth embodiment of the present invention.

The document search device shown in FIG. 17 is composed of the input unit 11, the display unit 12, the control unit 13, the document file 1 shown in FIG.
4, an input unit 4 corresponding to the word database 15, the search unit 16, the search result classifying unit 17, and the search result reconstructing unit 18.
1, display unit 42, control unit 43, document file 44, word database 45, search unit 46, search result classification unit 47,
It has a search result reconstruction unit 48. The feature of the document search device in FIG. 17 is that, in addition to the above-described components, attention is paid to the upper / lower relationship (upward / downward relationship) of words in the word database 45, and the search result reconstruction unit 48 performs reconstruction. A relationship selection unit 4 for selecting only a search key (a search result based on a user-specified relationship) with respect to a user-specified word
9.

Now, in the document search apparatus having the configuration shown in FIG. 17, the relation selection unit 49 uses the search result reconstructing unit 48 to reconstruct the search results classified by the search result classifying unit 47 and arrange them on the map. Then, only search results using search keys having a user-specified relationship with a user-specified word (main word) are selected as the target search results. Here, the word specified by the user is, in the initial state,
The word is input as a search key from the input unit 41 according to a user operation.

The control unit 43 arranges only the reconstructed result of the retrieval result reconstructing unit 48, that is, only the retrieval key (the retrieval result based on the user-specified relationship) with respect to the user-specified word (the central word) The displayed map is presented to the user through the display unit 42. At that time, the control unit 43 also presents a word of a higher concept or a lower concept of each search key together with the search key (search word). Here, a search key other than the relationship specified by the user (search result by) is not presented as a search result. It should be noted that a search key (a search result based on a search key) other than the relation specified by the user may be presented by changing the modification of a color, a frame line, or the like.

Further, the control unit 43 determines whether a word of a higher concept or a lower concept of the search key (search word) presented corresponding to a certain search key (search word) is operated by a user (for example, When the word is selected and specified from the input unit 41 by a mouse operation), the search result reconstructing unit 48 performs the reconstructing process again by using the word as a user-specified word (a central word). At this time, the upper / lower relation selecting unit 49 selects, from among the search results classified by the search result classifying unit 47,
With respect to a new user-specified word (main word), only search keys (search results based on the user-specified relationship) are selected.

Next, details of the processing performed by the search result reconstruction unit 48 and the relationship selection unit 49 will be described with reference to the flowchart of FIG. First, the results searched by the search unit 46 are classified by the search result classifying unit 47 for each search key. The search result reconstructing unit 48 is, of the search results classified by the search result classifying unit 47, a search key that is currently focused on, that is, a central search key (in the initial state, input from the input unit 41 according to a user operation. search key)
Is first arranged on a map (search result map) including the search key (step E1).

Next, the search result reconstructing unit 48 checks whether or not the processing has been completed for all the search results (step E2). Follow (Step E
3). Then, the search result reconstructing unit 48 checks whether or not the searched word exists as a search key in the search result (step E4).
Pass control to.

Then, the relationship selecting section 49 checks the relationship between the search key in the search result and the central search key (step E5). If the relationship between the two is a user-specified relationship (step E6), the search result reconstruction unit 48
The retrieved search key is arranged together with the search result on the map such that the relation with the already arranged search key on the map is the same as the network structure of the word database 45 (step E7). On the other hand, if the relationship between the two is not a user-specified relationship, that is, if the relationship between the search key in the search result that has been followed and the search key currently centered is not a user-specified relationship. Then, the search result reconstructing unit 28 does not arrange the search key in the search result of the search destination on the map, and returns to step E2.

Search result reconstructing section 48 and relation selecting section 49
Repeats the above processing (steps E3 to E7) until the processing is completed for all search results (step E2). The control unit 43 displays the reconstructed result of the search result reconstructing unit 48 performed in cooperation with the relationship selecting unit 49, that is, the map in which only the search results by the search keys having the user-specified relationship are arranged. 42 to the user.

In this manner, the presentation example (map display example) of the retrieval result by the retrieval key having the user-specified relation, which is presented to the user, specifies only the superordinate / subordinate concepts and synonyms as the user-specified relation. About when
As shown in FIG.

FIG. 19A shows an example of presenting a search result when "Beelzebub" is input as a search key (search word) and the input search key is the center. Here, the search key "Belzebub" entered first,
Only the upper / lower concepts specified by the user and the search keys related to only the synonyms are presented. Since search keys having other relationships are presented for each search key, the range of the result presentation can be changed by selecting it.

FIG. 19 (b) shows that the search results to be presented are not limited to the ones specified by the user, but are also presented other than the ones specified by the user. This is an example in which modifications such as are changed. In FIG. 19A,
Only the relations designated by the user are presented, and those hidden and invisible are presented here, but the effect is the same.

As described above, by designating only the relationship desired by the user, the desired search result can be efficiently presented. The first to fourth described above
The embodiment relates to the presentation of search results (centering on a search key) when performing a document search using a neural network type word database, but the method of presenting results applied in each embodiment is not necessarily It is not necessary that the present invention is fixed independently, and it is also possible to adopt a configuration capable of presenting a result obtained by combining them, or a configuration capable of selecting a presentation method by a user's instruction.

Further, in the first to fourth embodiments, when the search key is input as a word, the description has been given of the document search apparatus which expands and searches the search key and presents the search result. In addition to simple word input, even if the input is composed of words, such as a compound word or a sentence, the input is analyzed and divided into word units, and each word is used as a search key for document search. It can also be applied to devices. In this case, each of the divided words is expanded using the word database.

The processing procedure in the document search apparatus applied in each of the first to fourth embodiments is such that a program for causing a computer such as a personal computer capable of reading the program to execute the processing procedure is recorded. This is realized by causing a computer to read and execute a program recorded on a recording medium such as a CD-ROM, a DVD-ROM, a floppy disk, and a memory card. Here, the program recorded on the recording medium may be downloaded to a computer via a communication line or the like.

[0076]

As described above in detail, according to the present invention, words are connected in a network type, and a search key is developed using a neural network type word database capable of defining the strength of connection and the like. Thereby, related words such as complicated related words can be obtained as search keys from a search key specified by the user, and a more suitable search result can be obtained by performing a document search using this search key.

Further, according to the present invention, the relevancy of the search results is presented to the user in a manner that is easy to understand by presenting the results of the search using the neural network type word database in a network-type map that is intuitively understandable. Can be presented.

Further, according to the present invention, in a neural network type word database, attention is paid to the hierarchical relationship having a conceptual hierarchy, and only the relationships at the same level are expressed, whereby the word serving as a search key is obtained. Present each relationship clearly.

Further, according to the present invention, the relation at the same level is distinguished from the others, so that the relation for each word serving as a search key can be presented in an easy-to-understand manner. Further, according to the present invention, a search result desired by the user can be efficiently presented by selectively presenting a search key having a relationship designated by the user in the neural network type word database.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of a document search device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing an internal configuration of a search unit 16 in FIG.

FIG. 3 is a schematic diagram showing a data structure example of a word database 16 in FIG. 1;

FIG. 4 is a flowchart showing a processing procedure of a search key expanding unit 17 in FIG. 1;

FIG. 5 is a diagram showing an example of the contents of a search key list that is a search key expansion result when an input search word (search key) is “Belzebub”.

FIG. 6 is a flowchart showing a processing procedure of a search result reconstruction unit 18 in FIG. 1;

FIG. 7 is a diagram showing an example of a search result when “Beelzebub” is input as a search key (search word).

FIG. 8 shows a search result classified by a search result classifying unit 17 and a search result reconstructing unit 18 using the word database 1;
FIG. 13 is a diagram illustrating a map display example (example of presenting a search result) when reconfiguring according to the network structure of FIG.

FIG. 9 is a diagram showing an example of a search result for a search key (search term) “angel”.

FIG. 10 is a view showing a search result presentation example in which a search key and details of a search result using the search key are presented together.

FIG. 11 is a block diagram showing a schematic configuration of a document search device according to a second embodiment of the present invention.

12 is a flowchart showing a processing procedure of a search result reconstruction unit 28 and a hierarchical relation selection unit 29 in FIG. 11;

FIG. 13 is a diagram showing a presentation example (map display example) of a search result using a search key on the same hierarchy.

FIG. 14 is a block diagram showing a schematic configuration of a document search device according to a third embodiment of the present invention.

FIG. 15 is a flowchart showing a processing procedure of a search result reconstruction unit 38 and a hierarchical relation classification unit 39 in FIG. 14;

FIG. 16 is a table showing the relationship between a search result obtained by inputting a search key “Belzebub” and a search result obtained by using another search key when “Belzebub” is input as a search key. The figure which shows the example of presentation of the search result presented with a different frame line according to the upper-lower relationship with respect to "Ruzebub".

FIG. 17 is a block diagram showing a schematic configuration of a document search device according to a fourth embodiment of the present invention.

18 is a flowchart showing a processing procedure of a search result reconstruction unit 48 and a relationship selection unit 49 in FIG.

FIG. 19 is a diagram showing an example of presenting search results using a search key having a user-specified relationship (an example of a map display) in a case where only the upper / lower concepts and synonyms are designated as user-specified relationships.

[Explanation of symbols]

 11, 21, 31, 41 ... input unit 12, 22, 32, 42 ... display unit 13, 23, 33, 43 ... control unit 14, 24, 34, 44 ... document file 15, 25, 35, 44 ... word database 16, 26, 36, 46 ... search unit 17, 27, 37, 47 ... search result classification unit 18, 28, 38, 48 ... search result reconstruction unit 29 ... vertical relationship selection unit 39 ... vertical relationship classification unit 49 ... relationship Selection unit 161 Search control unit 162 Search key development unit 163 Document search unit

Claims (9)

[Claims] 1. A method in which words are connected in a network form with reference to a neural network type word database having the direction, type and strength of connection for each connection,
Performing a search key expansion process for expanding the search key based on the strength of the connection with the search key specified by the user; and searching the document using the search key expanded by the search key expansion process to obtain a search result And a step of presenting search results based on the configuration of the word database, and presenting the search results to the user while maintaining the relationship between the search results using a network-type map that maintains the same relationship as the network structure of the word database. And a document search method. 2. An input means for inputting a search key, a neural network type word database having words connected to each other in a network form, and having a direction, kind and strength of connection for each connection, A search unit that expands a search key based on a connection strength with a search key manually input by the input unit with reference to a database, and searches for a document using the expanded search key; and a search by the search unit. A search result classifying means for classifying results by a search key for each word of the word database; and a search result classified by the search result classifying means,
A search result reconstructing unit configured to reconstruct a network-type map based on the relation between words in the word database, and an output unit configured to present a user with a result of the reconstruction by the search result reconstructing unit. A document search device characterized by the following. 3. Referring to a neural network type word database having words connected to each other in a network and having a direction, kind and strength of connection for each connection,
Performing a search key expansion process for expanding the search key based on the strength of the connection with the search key specified by the user; and searching the document using the search key expanded by the search key expansion process to obtain a search result And presenting a search result based on the configuration of the word database. The hierarchical relationship of the word database on the network structure is regarded as a hierarchy, and the word is on the same hierarchy as the word specified by the user. Selecting a search result by a search key as a presentation target, and presenting the search result to a user while maintaining a relationship between the search results with a network-type map that maintains the same relationship as the network structure of the word database. Document search method. 4. An input means for inputting a search key, a neural network type word database connecting words to each other in a network form, and having a direction, kind and strength of connection for each connection; A search unit that expands a search key based on a connection strength with a search key manually input by the input unit with reference to a database, and searches for a document using the expanded search key; and a search by the search unit. A search result classifying means for classifying results by a search key for each word of the word database; and a search result classified by the search result classifying means,
A search result reconstructing means for reconstructing into a network type map based on the relation of the words in the word database, and as a target of reconstruction by the search result reconstructing means, in a hierarchical relationship of the words on the word database. A hierarchical relation selecting means for selecting a search result by a search key on the same hierarchy as a word specified by the user; and an output means for presenting a user with the reconstructed result by the retrieving result reconstructing means. Characteristic document search device. 5. A method in which words are connected to each other in a network and a neural network type word database having the direction, type and strength of connection for each connection is referred to.
Performing a search key expansion process for expanding the search key based on the strength of the connection with the search key specified by the user; and searching the document using the search key expanded by the search key expansion process to obtain a search result And presenting a search result based on the configuration of the word database. The hierarchical relationship of the word database on the network structure is regarded as a hierarchy, and a search key on the hierarchy is determined for each hierarchy. Presenting the result to the user while maintaining the relationship between the search results with a network-type map that maintains the same relationship as the network structure of the word database, with the result of the search being given a unique modification to the hierarchy. The document search method you want. 6. An input means for inputting a search key, a word database of a neural network type having words connected to each other in a network form, and having a direction, kind and strength of connection for each connection, A search unit that expands a search key based on a connection strength with a search key manually input by the input unit with reference to a database, and searches for a document using the expanded search key; and a search by the search unit. A search result classifying means for classifying results by a search key for each word of the word database; and a search result classified by the search result classifying means,
A search result reconstructing unit that reconstructs a network-type map based on the association of words in the word database, and a search result classified by the search result classifying unit,
Hierarchical relation classifying means for classifying each word in the hierarchical relation of words on the word database, and a modification result obtained by the search result reconstructing means, and a modification unique to each hierarchy according to the classification result by the hierarchical relation classifying means. Output means for presenting to a user with a tag attached thereto. 7. Words are connected in a network between words, and a neural network type word database having the direction, type and strength of connection for each connection is referred to.
Performing a search key expansion process for expanding the search key based on the strength of the connection with the search key specified by the user; and searching the document using the search key expanded by the search key expansion process to obtain a search result And a step of presenting a search result based on the configuration of the word database, wherein a search result using a search key having a user-specified relationship with a user-specified word is selected as a presentation target, and the word database is selected. And presenting the search results to the user while maintaining the relationship between the search results using a network-type map that maintains the same relationship as the network structure of the document. 8. An input means for inputting a search key, a word database of a neural network type having words connected to each other in a network and having a direction, a kind and a strength of connection for each connection; A search unit that expands a search key based on a connection strength with a search key manually input by the input unit with reference to a database, and searches for a document using the expanded search key; and a search by the search unit. A search result classifying means for classifying results by a search key for each word of the word database; and a search result classified by the search result classifying means,
A search result reconstructing means for reconstructing a network-type map based on the relation of the words in the word database; and a user in the word relation on the word database as a target of reconstruction by the search result reconstructing means. A relation selection unit for selecting a search result by a search key having a user-specified relationship with respect to a specified word; and an output unit for presenting a user with a result of the reconstruction by the search result reconstructing unit. Characteristic document search device. 9. Referring to a neural network type word database having words connected in a network and having a direction, a kind and a strength of connection for each connection,
Performing search key expansion processing for expanding the search key based on the strength of the connection with the search key specified by the user; and causing the search key to expand the search key and search the document using the search key expanded by the search key expansion processing. Acquiring and presenting search results based on the configuration of the word database, and presenting the search results to the user while maintaining the relationship between the search results in a network-type map that maintains the same relationship as the network structure of the word database. And a computer-readable recording medium recording a program to be executed by a computer.