JPH0991314A - Information search device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the information search device which classified information in conformity with a retrieved text and by a plurality of viewpoints that a user selects. SOLUTION: An information retrieval part 3 performs retrieval from an information storage part 1 according to a retrieval request that the user inputs to a retrieval request input part 2. A modification extraction part 4 analyzes modulation relation, regarding a word or phrase being requested to be retrieved from the text, in the retrieval result of the information retrieval part 3. A modification information group is stored in a storage part 5 together with the retrieved information and sent out to an output part 6. The user inputs relation that the word or phrase requested to be retrieved possibly has with other words and phrases as a viewpoint for classification from a viewpoint input part 7. An information classification part 8 extracts words and phrases having the relation inputted from the viewpoint input part 7 from the analysis result and performs classification by the extracted words and phrases. The classification results are stored in a storage part 5 and sent out to the output part 6.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information search apparatus, and more particularly to an apparatus for classifying and displaying retrieved information.

[0002]

2. Description of the Related Art Conventionally, when a large amount of digitized documents are stored and desired information is searched, a search is performed by designating search conditions, and whether the resulting document is necessary or not is read. In general, whether or not to confirm is determined by checking the contents of each document. However, this method may be used if the document to be sought is clearly known.For example, search for a document that is likely to be written about the content that you want to know vaguely, or what kind of document is related to a certain matter If the user wants to know the outline of the document, the work of confirming the contents of the document requires a lot of work. In such a case, if the search results are arranged and displayed, it is easy to grasp the outline of the search results. Further, in organizing, it is desirable that the organizing is not based on a fixed manner of organizing, but on the basis of the viewpoint from which the user views the information.

In response to such a request, there is a technique for displaying the search results by prioritizing them based on the degree of association between the search conditions and the search results. Good, but not enough to see the whole thing.

On the other hand, even if the desired information is ambiguous, in order to efficiently search for the desired information from a vast amount of documents, the results are not returned once by the search, but rather the results are returned in a single search, and the information is narrowed down little by little while looking at the overview. The way to go is desirable. As such a thing, for example, JP-A-6-215036.
There is a "Document Collection Search Method" described in Japanese Patent Publication. It divides documents into document sets without entering search criteria such as keywords, presents a summary of each set, allows the user to select one or more sets, and for the selected document set, Further division and presentation of the summary are repeated. However, even if it is ambiguous, the user has some clues to the desired information, and it is not efficient to classify documents without using them at all, and the classification is based on the viewpoint seen by the user. desirable.

As a means for displaying and organizing search results,
There is a "multi-dimensional browsing method" described in Japanese Patent Publication No. 5-24551. This technique allows the user to select a viewpoint by organizing search results from a plurality of viewpoints. However, a classification axis based on a plurality of viewpoints is prepared in advance, and it is also determined in advance which document falls into which classification when classification is performed using a certain classification axis. However, when the classification axis is prepared in advance, the prepared classification axis tends to be general. Further, when the stored documents cover various fields, formats and contents, it is difficult to prepare the classification axis. Furthermore, if many classification axes are prepared, the burden of preprocessing increases.

[0006]

The present invention has been made in view of the above-mentioned circumstances, and eliminates the need to prepare a classification axis in advance or perform classification according to the classification axis, and to retrieve the searched text. In accordance with the above, the information is categorized from the viewpoint desired by the user, and the desired information is searched while narrowing down the results by selecting the classification results and finely categorizing, and redoing the categorization if necessary. It is an object of the present invention to provide an information searching device that

[0007]

According to a first aspect of the present invention, in an information search apparatus, an information storage means for storing information including at least text information, a search request input means for inputting a search request, An information search unit that searches for information that includes the search request in the text information, and a word or phrase of the search request as a viewpoint for classifying the information searched by the information search unit and another word or phrase in the text information. A viewpoint input means for inputting a possible relationship and a textual information of each information searched by the information searching means, and a word having the relationship with the search request is extracted and searched based on a difference between the extracted words. An information classifying unit that classifies the information, a storage unit that stores the information searched by the information searching unit, and a classification result by the information classifying unit; It is characterized in that it has an output means for outputting the information and the classification result.

According to a second aspect of the present invention, in the information search apparatus, information storage means for storing information including at least text information, search request input means for inputting a phrase as a search request, and the search request input. Information retrieval means for retrieving information including the word / phrase input as a search request by means in the text information, and a relationship that the word / phrase may have with other words / phrases in the text information as a viewpoint for classifying the information And a viewpoint input means for inputting, and a phrase extracted by extracting a phrase having the relation with the phrase in the text information of each information of the information group forming all or a part of the information retrieved by the information retrieval device. An information classifying unit that classifies the information in the information group based on the difference, and one or more groups are selected from the results classified by the information classifying unit. Classification result selecting means, storage means for storing the information searched by the information searching means, the classification result by the information classifying means and the selection result by the classification result selecting means, and the searched information and the classification result. Classification by the information classification means based on the relationship input by the viewpoint input means, with the information belonging to one or more groups selected by the classification result selection means as an information group It is characterized by enabling.

According to a third aspect of the present invention, in the information search apparatus, a search for inputting information storage means for storing information including at least text information and two expressions as a search request and an expression representing a relationship between them. Request input means,
Information retrieval means for retrieving the information having an expression satisfying the retrieval request in the text information, and designating one of the two terms as a viewpoint for classifying the information retrieved by the information retrieval means And a viewpoint input means for designating a relationship different from the one used in the search among the relationships that the specified phrase may have with other words, and the viewpoint in the text information of each information searched by the information searching means. An information classifying unit that classifies the searched information based on the difference between the extracted words and phrases having the relationship specified by the viewpoint inputting unit, and the information searching unit Storage means for storing the retrieved information and the classification result by the information classification means, and output means for outputting the retrieved information and the classification result It is an butterfly.

According to a fourth aspect of the present invention, in the information searching apparatus, an information storing means for storing at least information including text information and a search request for inputting two words and phrases and an expression representing a relationship between them. Request input means,
An information retrieval means for retrieving the information having an expression satisfying the retrieval request in the text information, and one of the two terms as a viewpoint for classifying the information, and the designated term is Point-of-view input means for designating a possible relationship with another word and phrase, and a word or phrase having the above-mentioned relationship with the word or phrase in the text information of each information of the information group constituting all or part of the information searched by the information search means An information classifying unit that classifies the information in the information group based on a difference between the extracted words and phrases, and a classification result selecting unit that selects one or a plurality of groups from the classification results classified by the information classifying unit. A storage for storing the information retrieved by the information retrieval means, the classification result by the information classification means, and the selection result by the classification result selection means And a means for outputting the searched information and the classification result, the information belonging to one or more groups selected by the classification result selecting means as an information group, and further by the viewpoint input means. It is characterized in that classification can be performed by the information classification means based on the input relationship.

According to a fifth aspect of the present invention, in the information search apparatus, there is provided between an information storage unit that stores information including at least text information and a phrase of the text information stored in the information storage unit. In the analysis result storage means, an analysis result storage means for storing the result of analyzing the relationship, a thesaurus holding portion for holding a thesaurus showing a semantic vertical relationship of words, and text information including an input clue word Initial set selecting means for searching based on the analysis result of (1) and selecting as information forming a search space, search space holding means for holding the text information searched by the initial set selecting means in a searchable manner, and the clue word A viewpoint extracting means for extracting a plurality of possible relationships with other words in the text information, and the clue word corresponding to the viewpoint input in the text information. Hierarchical means for hierarchizing related terms based on the thesaurus held in the thesaurus holding unit, grouping means for grouping into one or more groups based on the hierarchical structure, and visualization for visualizing the groups. It is characterized by having means.

According to a sixth aspect of the present invention, in the information search apparatus, an information storage unit that stores at least information including text information, and a thesaurus holding unit that holds a thesaurus showing a semantic vertical relationship of words. , Initial set selecting means for searching the information storage means for text information including the input clue word and selecting it as information constituting a search space, and holding the text information searched by the initial set selecting means searchable A search space holding means, a dependency extraction means for analyzing text information including the clue word and extracting a dependency relationship of the clue word,
Based on the dependency relationship extracted by the dependency extraction means, a viewpoint extraction means for extracting a plurality of relationships in which the clue word / phrase can have other terms in the text information, and the clue word / phrase are input in the text information. Hierarchizing means for hierarchizing words having a relationship corresponding to the viewpoint based on the thesaurus held in the thesaurus holding part, and grouping means for grouping into one or more groups based on the hierarchical structure, It is characterized by having a visualization means for visualizing a group.

The invention according to claim 7 is the information searching apparatus according to claim 5 or 6, wherein the clue word and phrase is expanded using a thesaurus held in the thesaurus holding unit. It is characterized by being a phrase.

The invention according to claim 8 is the information search apparatus according to claim 7, wherein the clue word is only a part of the expanded word according to a user's instruction. It is a thing.

According to a ninth aspect of the present invention, in the information searching apparatus according to the fifth or sixth aspect, the grouping unit groups one or more groups instructed by a user in a lower hierarchy, and the visualization is performed. It is characterized by being visualized by means.

According to a tenth aspect of the present invention, in the information searching apparatus according to the fifth or sixth aspect, when the user inputs a new viewpoint, the hierarchizing means performs hierarchization based on the new viewpoint. It is characterized by performing.

The invention according to claim 11 is the information searching apparatus according to claim 5 or 6, wherein the hierarchizing means specifies when the user specifies one or more groups and inputs a new viewpoint. It is characterized in that the created groups are hierarchized based on the new viewpoint.

According to a twelfth aspect of the present invention, in the information searching apparatus according to the fifth aspect, when the user specifies a new clue phrase, the new clue phrase is included in the text information. Extracting a plurality of possible relationships with other words, the hierarchizing means holds the thesaurus holding unit with words having a relationship corresponding to the viewpoint in which the new clue word is input in the text information. It is characterized by being hierarchized based on.

According to a thirteenth aspect of the present invention, in the information search apparatus according to the sixth aspect, the dependency extraction means is
When the user specifies a new clue word, the text information including the new clue word is analyzed to extract the dependency relationship of the new clue word, and the viewpoint extracting means is extracted by the dependency extracting means. Based on the dependency relationship, the new clue word / phrase extracts a plurality of possible relationships with other words / phrases in the text information, and the hierarchizing means inputs the new clue word / phrase in the text information. It is characterized in that words and phrases having a relationship corresponding to a viewpoint are hierarchized based on the thesaurus held in the thesaurus holding unit.

[0020]

According to the invention described in claim 1, a search request,
As a viewpoint for classifying the retrieved information, the relation that the phrase of the search request may have with other phrases in the text information is input. Then, the information including the search request in the text information is searched from the information storage unit, and in the text information of each searched information, the word having the relationship input with the word of the search request is extracted, and the extracted word The retrieved information is classified and output based on the difference. This classifies the meaning and context of the search request word in the information searched by the search request word, based on the user's desired viewpoint, that is, the relationship between the input words. Can be output. By utilizing this, the user can easily search for desired information from a vast amount of information.

According to the second aspect of the present invention, as a viewpoint for classifying the search request and the searched information, a relation in which the phrase of the search request may have with other phrases in the text information is input. . Then, the information including the search request in the text information is searched from the information storage unit, and in the text information of each searched information, the word having the relationship input with the word of the search request is extracted, and the extracted word The retrieved information is classified and output based on the difference. The user can refer to the classification result, select a group of retrieved information, and input a new relationship of terms as a viewpoint, thereby classifying the classified group from another viewpoint. As a result, the user can narrow down the target information from the retrieved information, and can more easily find the desired information.

According to the third aspect of the present invention, an expression representing two phrases and the relationship between them is input as a search request, and the information satisfying this search request is searched for in the text information. . In addition, as a viewpoint for classifying the retrieved information, either one of the two terms input as the retrieval request is designated, and the relation that the designated term may have with other terms is used for retrieval. A different relationship than was specified. From the text information of each searched information, a word having a specified relationship with a specified word is extracted, and the searched information is classified and output based on the difference between the extracted words. As a result, only the information satisfying the expression input as the search request is searched, and the search results can be classified from another viewpoint.
Therefore, the user can easily search for desired information.

According to the fourth aspect of the present invention, an expression representing two words and a relation between them is input as a search request, and information in which the expression satisfying the search request exists in the text information is searched. . In addition, as a viewpoint for classifying the retrieved information, either one of the two terms input as the retrieval request is designated, and the relation that the designated term may have with other terms is used for retrieval. A different relationship than was specified. From the text information of each searched information, a word having a specified relationship with a specified word is extracted, and the searched information is classified based on the difference between the extracted words. The user refers to the classification result, selects a group of retrieved information, and inputs a word or phrase to be a new viewpoint and a relationship with the word or phrase, thereby classifying the classified group from another viewpoint. be able to. This will allow the user to
Target information can be narrowed down, and desired information can be searched more easily.

According to the fifth aspect of the invention, the relationship between the words and phrases of the text information stored in the information storage means is analyzed in advance and stored in the analysis result storage means. When the clue word is input, the initial set selection means searches the text information including the clue word based on the analysis result in the analysis result storage means, and the searched text information is used as the search space in the search space holding means. Hold on. The point-of-view extraction means extracts the relationship that the clue word / phrase may have with other words / phrases in the text information in the search space, and the layering means extracts the word / phrase having the relationship that the clue word / phrase corresponds to the viewpoint input in the text information. Layer based on the thesaurus. Then, the grouping means groups them into one or more groups based on the hierarchical structure, and the visualization means visualizes them.

According to the invention described in claim 6, the initial set selection means searches the information storage means for text information including the input clue word, and uses the searched text information as a search space. Hold in holding means.
The dependency extraction means analyzes the text information including the clue word and extracts the dependency relationship of the clue word,
The viewpoint extraction means extracts a plurality of relationships that the clue word / phrase can have with other words / phrases in the text information, based on the dependency relationship extracted by the dependency dependence extraction means, and the hierarchization means extracts the clue word / phrase in the text information. The words and phrases that have a relationship corresponding to the viewpoint input in step 2 are hierarchized based on the thesaurus. Then, the grouping means groups them into one or more groups based on the hierarchical structure, and the visualization means visualizes them.

According to the inventions of claims 5 and 6, as in claim 1, the user desires what meaning and context the clue word / phrase is used in the information retrieved by the clue word / phrase. It can be grouped and output based on the viewpoint. By utilizing this, the user can easily search for desired information from a vast amount of information.

As for the clue word, as in the invention described in claim 7, the word entered by the user is expanded using the thesaurus held in the thesaurus holding unit, and is searched by the word of the subordinate concept. Can be configured. At this time, as in the invention described in claim 8, it is possible to use only a part of the expanded phrases according to the instruction of the user as the clue phrase.

According to the invention described in claim 9, information can be narrowed down by grouping and visualizing one or more groups instructed by the user in a lower hierarchy. Further, according to the invention described in claim 10, when the user inputs a new viewpoint, hierarchization is performed based on the new viewpoint, and visualization is performed by grouping, so that different viewpoints can be displayed in the same search space. Classification can be performed. At this time, if the user specifies one or more groups and inputs a new viewpoint as in the invention described in claim 11, the details of the specified group are classified based on the new viewpoint, Can be displayed. Further, according to the invention as set forth in claims 12 and 13, when a new clue word is input from the user, hierarchization and grouping are performed based on the new clue word within the same search space. , Can be visualized. As described above, according to the ninth to thirteenth inventions, new clue words and phrases are classified and narrowed down by a new clue word in the search space including the text information searched by the clue word first input by the user. Even if the search intention is not clear, desired information can be easily searched from a vast amount of information, and the search can be performed while grasping the whole.

[0029]

1 is a block diagram showing a first embodiment of an information searching apparatus of the present invention. In the figure, 1 is an information storage unit, 2 is a search request input unit, 3 is a search unit, 4 is a dependency extraction unit, 5 is a storage unit, 6 is an output unit, 7 is a viewpoint input unit,
Reference numeral 8 is an information classification unit.

The information storage unit 1 is a storage device that stores a large amount of information. The information storage unit 1 may have any configuration as long as it can store information, such as a memory, a magnetic disk, an optical disk, or the like. Further, the stored information may include a picture or a figure as long as it includes at least text.

The search request input unit 2 is composed of input means such as a keyboard and a mouse. The user
The word W is input as a search request from the search request input unit 2.

The information search unit 3 searches the information storage unit 1 using the word W input by the search request input unit 2 as a search key. As a search method, a keyword is extracted in advance from the information stored in the information storage unit 1 and the keyword and position information are registered in the index table, a full-text search method in which a text is directly searched, and the like. Any method may be used, and various search techniques conventionally used can be used.

The dependency extraction unit 4 analyzes a part of the text of each information obtained as a result of the search by the information search unit 3,
The dependency relationship of the word W used as the search key is extracted, and this dependency information is stored in the storage unit 5 together with the retrieved information. The part of the text is a range in which dependency analysis is performed centering on the word W, and is analyzed in a predetermined range such as one sentence, two sentences, and one paragraph. Of course, the entire text may be parsed if desired. For analysis, various conventionally used analysis methods can be used.

The storage unit 5 stores the retrieved information and the dependency information passed from the dependency extraction unit 4, and
The classification result passed from the information classification unit 8 is stored.

The output unit 6 is composed of, for example, an output device such as a display, and outputs the search result to the output device. The output unit 6 outputs the title information of the search result, or outputs the whole sentence if necessary. The output unit 6 also outputs the classification result of the information classification unit 8 to the output device.

The viewpoint input section 7 is composed of input means such as a keyboard and a mouse. These may be common to the search request input unit 2. The user inputs from the viewpoint input unit 7 the relation REL1 in which the word W can have other words in the text as a viewpoint for classifying the search results. The viewpoint is input by, for example, extracting a relationship that the word W can have with other words from the dependency information, displaying the relationship on the display, and designating the relationship from the user by using the keyboard or the mouse. be able to.

The information classifying unit 8 extracts the word / phrase W and the word / phrase having the relation REL1 input by the user in the viewpoint input unit 7 from the dependency information of the search result, and retrieves the search result based on the difference between the extracted words / phrases. Classify. The information classification unit 8 stores the classification result in the storage unit 5. Each time a new viewpoint is input to the viewpoint input unit 7, the information classification unit 8 performs reclassification according to the new viewpoint.

Next, the operation of the first embodiment of the present invention will be described using a specific example. FIG. 2 shows the first of the present invention.
FIG. 7 is an explanatory diagram showing an example of an operation and an intermediate process when “release” is input as a search request in the embodiment of FIG.
It is assumed that the information storage unit 1 stores at least a text having the following parts. TEXT1: Company A announced that it will develop a workstation with high processing speed and will release it in July. TEXT2: Company B will release a workstation developed by receiving technical guidance from Company C at the end of the year. TEXT3: Company D will launch a new workstation next month, which was supplied by Company F, as the first of the alliance business with Company E. TEXT4: The workstation that Company B released three years ago is ... TEXT5: Company F has released a color display for workstations. ...

The user selects "release" as the word W of the search request.
Is input to the search request input unit 2, the search request input unit 2 sends it to the information search unit 3. The information retrieval unit 3 is "on sale".
Is used as a search key to search the information storage unit 1, and the search result is sent to the dependency extraction unit 4. TEXT1 to TE described above
XT5 is searched because it includes all the search keys "release". In addition to this, it is assumed that a lot of information is searched.

The dependency extraction unit 4 analyzes the dependency relationship of “release” in the text included in the search result of the information search unit 3. As a range of text analysis, here, only sentences including “release” are analyzed. The dependency relationship of the above-mentioned TEXT1 to TEXT5 is as follows. TREE1: (Released (Operator by Company A) (Target workstation (...)) (Time July)) TREE2: (Released (Operator by Company B) (Target workstation (...)) (Time End of year) ) TREE3: (Release (Operator: Company D) (Target workstation (...)) (Time next month)) TREE4: (Release (Operator: Company B) (Target workstation (...)) (Time 3) A year ago)) TREE5: (Launch (operator F company) (target display (use workstation) (color type)))

For example, TREE1 indicates that the operation subject of “release” is “company A”, the target is “workstation”, and the period is “July”. TREE1 ~
The dependency information group including the TREE 5 is stored in the storage unit 5 together with the retrieved information. The search result stored in the storage unit 5 is sent to the output unit 6, and, for example, title information of these pieces of information is displayed on the display.

When the search result is displayed, the user inputs a viewpoint for classification from the viewpoint input unit 7. Viewpoint input section 7
Is a sale in the dependency information as a selectable viewpoint.
Shows the relationships that has with other words and phrases. TREE1 to TRE
In the case of E5, "moving subject / target / time" is displayed as the selectable viewpoint. Here, the user selects how he / she wants to see the word having a relationship with “release”. For example, if you want to know who is selling,
Is selected by, for example, a mouse, the selection result is sent to the information classification unit 8.

The information classifying section 8 takes out a case element representing the "action subject" from the analysis result. TREE1 to TREE
In the case of 5, "Company A", "Company B", "Company D",
“Company B” and “Company F” are retrieved. The information classification unit 8
By comparing these, the same things are classified into the same group, and different things are classified into different groups. In this case, TREE2 and TREE4 are classified into the same group. The classification result is
For example, (("Company A" TREE1 ...) ("Company B" TR
EE2 TREE4 ...) ("D company" TREE3
...) ("Company F" TREE5 ...) ...) The information classification unit 8 stores these classification results in the storage unit 5.

The classification result stored in the storage unit 5 is sent to the output unit 6. The output unit 6 displays the classification result of the information classification unit 8 on the display. Various display methods can be considered. FIG. 3 is an explanatory diagram illustrating an example of a display of the classification result according to the first embodiment of this invention. In FIG. 3, each piece of information is indicated by a dot, and each piece of information is classified into “Company A” to “Company F” by the “action subject”. Here, when the user clicks an arbitrary point with, for example, a mouse, the title or the text of the information is displayed.

Here, when a viewpoint of another classification is input from the viewpoint input unit 7, the information classification unit 8 reclassifies the search results based on this new viewpoint. For example, when the viewpoint "time" is input here, it is newly classified based on the input and displayed on the display.

FIG. 4 is a block diagram showing the second embodiment of the information search system of the invention. In the figure, the same parts as those in FIG. Reference numeral 9 is a search request input unit, 10 is an information search unit, 11 is a search unit, 12 is a dependency extraction unit, 13 is a relationship comparison unit, and 14 is a viewpoint input unit.

The search request input unit 9 is composed of input means such as a keyboard and a mouse. The user
A search request in natural language is input from the search request input unit 9. The search request is an expression that expresses the relationship between the two words and phrases. The search request input unit 9 analyzes the search request input by the user and extracts the two terms W1 and W2 and the relation REL2 between them. Various conventionally used analysis methods can be used for the analysis. The search request input unit 9 sends the words W1 and W2 to the search unit 11,
The relation REL2 is sent to the relation comparison unit 13.

The information retrieval unit 10 is composed of a retrieval unit 11, a dependency extraction unit 12, and a relation comparison unit 13. Search unit 1
1 searches the information storage unit 1 on the condition of the logical product of the words W1 and W2. As a search method, a keyword is extracted in advance from the information stored in the information storage unit 1 and the keyword and position information are registered in the index table, a full-text search method in which a text is directly searched, and the like. Any method may be used, and various search techniques conventionally used can be used. Here, it is assumed that the text stored in the information storage unit 1 has not been touched in advance and the full text search method is used.
The search unit 11 sends, to the dependency extraction unit 12, information in which, of the texts that include both the words W1 and W2, both are within a certain range. The certain range may be within one sentence, within two sentences, within the same paragraph, or the like. Of course, such a range may not be defined, and all that are included in the text may be used as the search result.

The dependency extraction unit 12 analyzes a part of each text in the information output from the search unit 11 to obtain the word W.
The dependency relationship of 1 and W2 is checked, and the dependency information is sent to the relationship comparison unit 13. Part of the text is the word W
It is the range that the dependency of 1 and W2 has, and analyzes in a predetermined range such as one sentence, two sentences, and one paragraph. Of course, the entire text may be analyzed. Various conventionally used analysis methods can be used for the analysis.

The relation comparing unit 13 compares the relation between the words W1 and W2 in each dependency information with the relation RE of the analysis result of the search request.
Compare with L2. As a result, the dependency information in which both match is stored in the storage unit 5 together with the retrieved information.

The viewpoint input unit 14 is, for example, a keyboard,
It is composed of input means such as a mouse. You may use what is common with the search request input part 9. The user is the word W
Either 1 or W2 is designated, and as a viewpoint for classifying the search results, the relationship that the phrase may have with other phrases in the text is input. The viewpoint can be input, for example, by extracting a relationship in which the words W1 and W2 can have other words from the dependency information and displaying the extracted relationship, and the user can select a desired one from the relationships.

Next, the operation of the second embodiment of the present invention will be described using a specific example. FIG. 5 shows a second embodiment of the present invention.
FIG. 7 is an explanatory diagram showing an example of an operation and an intermediate process when “sale workstation” is input as a search request in the embodiment of the present invention. It is assumed that the information storage unit 1 stores at least TEXT1 to TEXT5, as in the first embodiment.

When the user inputs "sale workstation" as a search request, the search request input unit 9 analyzes this and extracts the following three elements. W1 =
Release, W2 = workstation, REL2 = target search request input unit 9 sends the words “release” and “workstation” to the search unit 11, and the relation “target” is compared to the relation comparison unit 1
Send to 3.

The search unit 11 searches the information storage unit 1 using the logical product of “release” and “workstation” as a search condition, and sends the search result to the dependency extraction unit 12. Here, it is assumed that the condition is satisfied only when two words and phrases as the search condition are present in one sentence. TEXT1 described above
~ TEXT5 is searched because it satisfies all the conditions. In addition to these pieces of information, it is assumed that a lot of information has been searched.

The dependency extraction unit 12 analyzes the dependency relationship existing between the words “release” and “workstation” in the text included in the search result of the search unit 11. As a range of text analysis, only sentences including “release” and “workstation” are analyzed here. The dependency relationship of the above-mentioned TEXT1 to TEXT5 is TREE1 to T like the first embodiment.
It becomes REE5, and these are sent to the relation comparison unit 13.

The relationship comparing unit 13 compares the relationship between “sales” and “workstation” in each dependency information with the relationship “target” of the analysis result of the search request. For example, TRE
In the case of E1, since the relationship between “release” and “workstation” is “target”, both match and it is determined that TREE1 satisfies the search request. On the other hand, in TREE5, since there is no direct relationship between "release" and "workstation", it is determined that TREE5 does not satisfy the search request. In this way, it is determined that TREE1 to TREE4 among TREE1 to TREE5 satisfy the search request, and the information group including these is stored in the storage unit 5 as the search result of the information search unit 10. The search result stored in the storage unit 5 is sent to the output unit 6, and, for example, title information of these pieces of information is displayed on the display.

When the search result is displayed, the user inputs a viewpoint for classification from the viewpoint input unit 14. The viewpoint input unit 14 displays, as selectable viewpoints, all the relationships related to the words “release” and “workstation” in the dependency information extracted by the dependency extraction unit 12, for example. Here, when the user selects, for example, the phrase “release” and the relationship “execution subject” of the phrase “release” with the mouse, the selection result is the information classification unit 8
Sent to

The information classifying unit 8 extracts a word representing "actor" of "release" from the dependency information. TREE
In the case of 1 to TREE4, "A company", "B company",
“Company D” and “Company B” are retrieved. The information classifying unit 8 compares these terms and classifies the same into the same group and different ones into different groups. In this case, they are classified as follows. (("Company A" TREE1 ...) ("Company B" TR
EE2 TREE4 ...) ("D company" TREE3
...) ...) The information classification unit 8 stores these classification results in the storage unit 5.

The classification result stored in the storage unit 5 is sent to the output unit 6, and the output unit 6 displays the classification result on the display. Here, when a viewpoint of another classification is input from the viewpoint input unit 14, the information classification unit 8 reclassifies the search results based on this new viewpoint.

FIG. 6 is a block diagram showing the third embodiment of the information search system of the invention. In the figure, those parts that are the same as those corresponding parts in FIG. 4 are designated by the same reference numerals, and a description thereof will be omitted. Reference numeral 15 is an information classification unit, and 16 is a thesaurus. In the third embodiment, an example in which information is classified using the thesaurus 16 is shown.

The information classifying unit 15 selects the word W1 or W selected by the viewpoint input unit 14 from the dependency information of the search result.
Words having a relationship input with 2 as a viewpoint are extracted, and similar ones are classified into the same group using the thesaurus 16. For the classification, for example, a level is determined in advance, and if the groups are equal to or higher than the level, they are classified into the same group.

The thesaurus 16 is a dictionary constructed hierarchically according to the inclusion relation of the concept of words. FIG. 7 is an explanatory diagram of an example of the thesaurus. For example, terms such as “workstation” and “personal computer” are included in the concept of the term “digital computer” as its concept. Therefore, the words "workstation" and "personal computer" are placed under the words "digital computer". Similarly, the phrase "digital computer"
The concept of is included in the concept of the word “computer”, and is placed under the word “computer”.

Next, an example of the operation of the third embodiment of the present invention will be described using a concrete example. Here, the operation in the case where “developed by X company” is input as the search request will be described. The search result is stored in the storage unit 5 after the search request is input.
The operation until it is stored in is similar to that of the second embodiment. That is, a search is performed using a logical product of the two phrases "X company" and "development" as the search condition, and further, the relationship "action subject" is established as the relationship between the phrase "X company" and "development".
Information having “” and dependency information corresponding to the information are stored in the storage unit 5.

Here, it is assumed that the dependency information of the search result stored in the storage unit 5 includes the following items. TREE11: (development (operator X company) (target)
Workstation)) TREE12: (Development (operator X company) (Target)
Computer)) TREE13: (Development (operator X company)) (Target)
PC)) TREE14: (Development (operator X company)) (Target)
Laptop)) TREE15: (Development (operator X company)) (Target)
PDA)) TREE16: (Development (operator X company) (Target)
Copier)) TREE17: (Development (operator X company)) (Target)
Facsimile)) Further, it is assumed that the thesaurus 16 stores the contents shown in FIG.

After the search results are displayed, the user selects the word “development” from the two words and phrases used as the search condition from the viewpoint input unit 7 as the classification viewpoint, and designates the “object” as the relationship. And The information classifying unit 15 refers to the thesaurus 16 and classifies words having a relation “object” with respect to the word “development” based on whether or not they are in the same group at a preset level or higher.

For example, in the thesaurus 16 shown in FIG. 7, when level 4 is set, the phrase "calculator" is used.
The terms such as "calculator" and "digital computer" that are subordinated to "," and the terms such as "workstation" and "personal computer" that are further subordinated are the terms included in the concept of the term "calculator". . Therefore, information having these terms as the relation “target” is classified into the same group. The above-mentioned TREE11 to TREE17 are
It is classified as follows. (("Calculator" TREE11 TREE12 TRE
E13 TREE14 TREE15) ("Copier" TREE16) ("Facsimile" TREE17)) Further, when level 6 is set, it is classified as follows. (("Workstation" TREE11) ("Computer" TREE12) ("PC" TREE13 TREE14) ("PDA" TREE15) ("Copier" TREE16) ("Facsimile" TREE17))

The classification result of the information classification unit 15 is stored in the storage unit 5. The classification result stored in the storage unit 5 is sent to the output unit 6 and displayed on the display. FIG. 8 is an explanatory diagram illustrating an example of a display of the classification result according to the third embodiment of this invention. Here, a display example when level 4 is set is shown. The information is classified and displayed for each level 4 phrase. In this way, even if the words and phrases used in the information are different, the information indicating the same concept can be classified and displayed in the same group.

The classification criteria are not limited to those in the above-mentioned embodiment, and various methods can be used. Instead of using the level that the thesaurus has in advance,
You may classify using the distance to a common parent. Further, the classification may be carried out according to a predetermined criterion, and if this does not fall within a certain range (for example, 10 to 20 pieces), the classification may be changed again and the number may fall within the range.
Further, the number of classifications may be adjusted by presenting the result of classification according to a predetermined standard to the user and giving an instruction to change the standard as necessary.

The thesaurus is not limited to the above,
Anything can be used, such as an existing one or one created for this information search device, and the number of levels and the like are arbitrary.

In the above-described third embodiment, the thesaurus 16 is applied to the second embodiment, but it can be applied to the first embodiment.

FIG. 9 is a block diagram showing the fourth embodiment of the information search system of the invention. In the figure, those parts that are the same as those corresponding parts in FIG. 4 are designated by the same reference numerals, and a description thereof will be omitted. Reference numeral 17 is a classification selection unit, and 18 is an information classification unit. In the above-described second embodiment, when the user inputs a new viewpoint after displaying the classification result, the entire search result is reclassified according to the new viewpoint. In the fourth embodiment, the user selects one or a plurality of groups according to his / her interest, for example, a group that is likely to be related to his / her own interest, from among several classified groups. When a new viewpoint is input, the information included in the selected group is targeted and classified by the newly input viewpoint.

The classification selection section 17 is composed of input means such as a mouse. You may use what is common with the search request | requirement input part 9 and the viewpoint input part 14. The user
One or a plurality of groups are selected from the groups classified by the information classifying unit 18.

The information classifying unit 18 extracts the word / phrase W and the word / phrase having the relation REL1 input by the user in the viewpoint input unit 7 from the dependency information of the search result, and retrieves the search result based on the difference between the extracted words / phrases. Classify. The information classification unit 8
The classification result is stored in the storage unit 5. The information classification unit 8
A group is selected from the classification selection unit 17, and the viewpoint input unit 1
When a new viewpoint is input to 4, the information included in the group selected by the classification selecting unit 17 is classified according to the new viewpoint, and the classification result is stored in the storage unit 5.

Next, an example of the operation of the fourth embodiment of the present invention will be described using a specific example. The operation up to the point where the user inputs the search request, performs the search, displays the search results, and then the user inputs the viewpoint of classification and the results classified according to the viewpoint are displayed is the same as the second embodiment described above. It is similar to the form.

The user selects one or a plurality of groups from the several classified groups in accordance with his / her interest, for example, a group likely to be related to his / her desire to know. After this,
When a new viewpoint is input from the viewpoint input unit 14, the information classification unit 18 classifies the group selected by the classification selection unit 17 according to the newly input viewpoint and stores it in the storage unit 5. . The storage unit 5 sends the classification result of the group to be classified to the output unit 6, and the output unit 6 displays this on the display.

For example, as described in the second embodiment, the phrase "release" is selected for the result retrieved by the retrieval request "release workstation" and the relation "operating subject" is designated. The result of the classification is displayed. At this point, the user selects a group whose operation subject is “Company B” as a classification target in the classification selection unit 17, and further inputs “time” as a new viewpoint from the viewpoint input unit 14. The group whose operating entity is “Company B” includes TEXT2 and TEXT4. By referring to these dependency relationships TREE2, TREE4, etc., the same "time" is classified as the same group, and different ones are classified as different groups. The classification result is as follows. ((“Company B” (“Year End” TREE2)) (“3 years ago”
TREE4) ・ ・ ・))

After that, by further selecting a group and repeating the classification from a new viewpoint as necessary, the user can easily find the information having the desired content.

In the description of the fourth embodiment, the classification selecting unit 17 and the like are added to the configuration of the second embodiment, but the configuration is not limited to this, and the first and third embodiments are also included. It can also be configured by adding to the embodiment.

In the above-described second to fourth embodiments, an example in which a verb and a case relationship between the verb and the case element are used as the search request has been described, but the present invention is not limited to this, and two concepts are used. Anything that represents the relationship between For example, a noun and its modifier may be classified as a viewpoint based on what attributes such as color and size are used for modification.

In each of the above-mentioned embodiments, the texts are analyzed to the extent necessary for classification at the time of retrieval and used for classification. However, the present invention is not limited to this, and is required each time during retrieval or classification. You may make it analyze a part.

Further, in the second to fourth embodiments, the method of storing the text which is not processed in advance and analyzing it at the time of retrieval to extract the dependency has been described.
The search method and the dependency extraction method are not limited to this, and various methods can be used. For example, it is possible to store the analyzed text in which the modification of the text is extracted in advance in the information storage unit 1 and use this to perform the search and the classification. Further, it is possible to use a text that has been subjected to a light analysis such as a morpheme analysis in advance to perform a further necessary analysis at the time of search or classification. Further, instead of referring to the entire text, a plurality of phrases or sentences representing the subject of the text may be extracted in advance, and this may be used as a target for searching and classification. In addition, the subject of the extracted text is subjected to dependency analysis and a subject index is added, and the subject index is used to search for text that satisfies the search request. Also, the subject index part has a specified relationship. You may comprise so that a phrase may be extracted.

In each of the above-described embodiments, the case where a natural language expression is input as a search request has been described, but the present invention is not limited to this, and a phrase or a relationship may be directly input.

In each of the above-described embodiments, the case where a search is performed when there is a phrase that matches the phrase input as the search request has been described. However, using a thesaurus or the like, synonyms and synonyms exist in the text. In some cases, the search may be considered to be satisfied and the search may be performed. At this time, the thesaurus used may be the same as the thesaurus 16 used in the third embodiment, and the thesaurus 16 can be shared in the third embodiment.

FIG. 10 is a block diagram showing the fifth embodiment of the information searching system of the invention. In the figure, 21 is an information storage unit, 22 is an analysis result storage unit, 23 is a search mechanism, 24 is an input unit, 25 is an output unit, 26 is a thesaurus holding unit, 31 is an initial set selecting unit, and 32 is a search space holding unit. , 33 is a viewpoint extraction unit, 34 is a layering unit, 35 is a grouping unit, and 36 is a visualization unit. In the fifth embodiment and a sixth embodiment to be described later, after classifying based on a certain viewpoint, subclassification is performed by enlarging a part, adding a viewpoint, changing an attention word, or changing the viewpoint. An example is shown in which the whole can be re-classified or the history can be manipulated to freely search. In the fifth embodiment, an example in which text information is analyzed in advance is shown, and in a sixth embodiment described later, an example in which text information is analyzed at the time of search is shown.

The information storage unit 21 has the same configuration as the information storage unit 1 in the above-described first embodiment and stores a large amount of information. The entity of the information does not necessarily have to be present here, and the information existing in a plurality of places such as on the network may be virtually managed. Further, the stored information may include a picture or a figure as long as it includes at least text.

The analysis result storage unit 22 stores the result of analyzing the relation between words and phrases in the text, with respect to the text information stored in the information storage unit. The analysis of the relationship may be performed by storing a dedicated analysis unit that has been analyzed and stored in advance, or may be stored after analysis of the relationship by some method. For the analysis of the relationship, various analysis methods that have been conventionally used can be used, and the analysis depth and the like can be arbitrarily set as needed.

FIG. 11 is an explanatory diagram of an example of the analysis result. Here, an example of the analysis result in the case of performing the morpheme analysis and the light analysis such as pattern matching is shown. By performing a morpheme analysis on the original sentence shown in FIG. 11 (A), it is decomposed into each morpheme separated by “/” in FIG. 11 (B). Then, the connection of the parts of speech of each morpheme is checked by pattern matching or the like, and the relationship between each morpheme is extracted, thereby
A relationship can be obtained with the word list as shown in (C). In the example shown in FIG. 11C, for example, “PD
A link indicates that there is a relation represented by the case particle "wo" between "A" and "development" + "done".

Returning to FIG. 10, the search mechanism 23 is a part for performing a search, and includes the initial set selection unit 31 and the search space holding unit 3.
2, viewpoint extraction unit 33, layering unit 34, grouping unit 3
5 and the visualization unit 36. Initial set selection unit 3
Reference numeral 1 is a part for selecting information to be used for search from the entire information. The information used for the search is a set including the text information obtained by searching the analysis result storage unit 22 with the words and phrases developed using the thesaurus provided by the thesaurus holding unit 26 based on the words and phrases given by the user. . The search space holding unit 32 holds the structure of the set including the text information obtained by the initial set selection unit 31 and information necessary for searching. When a certain phrase is given, the viewpoint extracting unit 33 extracts all the relationships that the phrase has with other phrases in the text to be searched, and creates a list of viewpoints based on this. The hierarchization unit 34 hierarchizes words and phrases having a relationship corresponding to the viewpoint in which the clue word is input in the text information, based on the thesaurus of the thesaurus holding unit 26. The grouping unit 35 groups the layers into one or more groups. The visualization unit 36 visualizes the grouped words and phrases as a diagram or the like.

The input section 24 is a section through which the user inputs data and gives instructions when searching for information.
For example, it is composed of input means such as a keyboard and a mouse. The output unit 25 is for presenting information to the user, and is composed of, for example, a display.

The thesaurus holding unit 26 holds the thesaurus. The thesaurus may be any one that systematizes the semantic hierarchical relationship between words. It may be a thesaurus created by itself, an existing thesaurus available in the form of a dictionary, or based on these. It may be created.

Next, an example of the operation of the fifth embodiment of the information search system of the invention will be described. FIG.
FIG. 14 is an explanatory diagram of an outline of an example of an operation in the fifth embodiment of the information search system of the invention. The text information stored in the information storage unit 21 is analyzed in advance, and the analysis result is stored in the analysis result storage unit 22. From the document set pre-analyzed in this way, first, a portion likely to be related to the content to be searched is selected as an initial set.
Then, the initial set is classified from the viewpoint specified by the user.
Then, based on the user's instruction, the user searches for information by narrowing down or re-classifying based on the classification.

This operation will be further described below. First,
The operation of selecting the initial set will be described. The user inputs a word or phrase representing a central concept of information to be searched for into the input unit 24 as a key for selecting the initial set. The user can input various items such as words, phrases and sentences as keys. Here, it is assumed that a word is input as a key.

When a word is input to the input section 24 as a key, it is sent to the initial set selection section 31. Initial set selection unit 3
1 expands a word input as a key using a thesaurus into synonyms and synonyms. The expanded words are
It is held in the search space holding unit as the central concept of the search. The initial set selection unit 31 retrieves a sentence including these words from the analysis result stored in the analysis result storage unit 22, and extracts the sentence analysis result and the ID of the information including the sentence as a sentence unit. The searched set of sentence units forms a search space as an initial set and is held in the search space holding unit 32.

As the search method, various conventionally used search techniques can be used. Here, it is assumed that the analysis result storage unit 22 has a word index based on the analysis result.

The expansion of the key word is such that, for example, when the word "PDA" is input, "PDA",
It is expanded to words such as "mobile terminal", "mobile information terminal", and "mobile information communication terminal". Here, the input word is expanded using the thesaurus, but of course, the search may be performed only by the input word.

When the user inputs a word or phrase as a key such as a phrase or a sentence, a method of extracting the independent words in the words and searching for a sentence including all of them, or the same independent word, Various methods such as a method for searching the same sentence can be applied to the relationship between independent words.

When the initial set is selected, next, the selected initial set is classified. The word input as a key is the word that becomes the center of the search, and is hereinafter referred to as the word of interest. In the search viewpoint extraction unit 33, the relationship that the target word has with other phrases in the text is extracted as a classification viewpoint.

FIG. 13 is an explanatory diagram of an example of the viewpoint conversion table. As shown in FIG. 13, the viewpoint conversion table is composed of words indicating the contents of viewpoints and corresponding case particles. For example, when the word "PDA" is searched for, and the sentence as shown in FIG. 11 is included as an initial set, by searching the link of the attention word "PDA", You can see that they have a relationship of "w". For example, this is converted into a meaning representing the relationship using a viewpoint conversion table as shown in FIG. 12, and is extracted as a name “target” that is easy to understand as a viewpoint. Similarly, viewpoints are extracted from all texts. Then, a list of viewpoints is created excluding the overlapping ones, and the output unit 25
To be displayed.

Here, the viewpoint is converted only by the correspondence between the viewpoint name and the case particle, but the conversion from such a relationship to the viewpoint is performed by using the meaning of a word or the like into a detailed meaning or context. The meaning may be extracted. Further, when a detailed relationship is extracted in advance by deep analysis, it may be used as a viewpoint as it is.

When the viewpoint list is displayed, the user
A point of view used for classification is selected and input from the input unit 24. For example, from the list of viewpoints displayed on the display, click the one you want to select with the mouse or the like.

The viewpoint selected by the user is sent to the hierarchizing unit 34. The hierarchizing unit 34 has a word having a relation expressed by a selected viewpoint with a word of interest (hereinafter, referred to as a related word) from all the sentences of the initial set forming the search space.
To extract. Furthermore, the semantic hierarchical relation between the extracted relational words is hierarchized using a thesaurus. Here, the frequency of appearance of the related words that overlap is held.
This hierarchical structure is held in the search space holding unit as the current search space.

FIG. 14 is an explanatory diagram of an example of hierarchical structure based on the thesaurus. FIG. 14 shows an example of hierarchical relational words when “means” is selected as the classification viewpoint in the example in which the initial set is selected by “PDA”. That is, the words "PDA" and the words "schedule management" and "business card management" that are related by the words "PD" and "mobile terminal", "mobile information terminal", "mobile information communication terminal", etc. , "FAX",
"Electronic mail", "PC communication", etc. are summarized by the thesaurus into words such as "management", "communication", etc.
It is hierarchized as shown in.

Then, the grouping unit 35 classifies the hierarchical structure of the related words into several groups. The number of classifications should be a predetermined number or a number within a predetermined range. Alternatively, the user may designate the number of groups to be classified here.

Various methods can be considered as the classification method. For example, when viewed from the root of the thesaurus, think that all groups have the same level of abstraction,
From the root of the hierarchical tree, look down from the root and divide by the level at which the number of groups is appropriate, or from the leaves of the hierarchical tree to the bottom up so that the number of members in all groups is almost the same. , Any method may be used, such as a method in which the number of groups is adjusted to an appropriate number, and the number may be changed depending on the purpose of the search.

The grouped hierarchical structure has a search space holding unit 3 as a grouping space, separately from the current search space.
2 is held.

The grouping space held in the search space holding unit 32 is visualized in the visualizing unit 36 and is output to the output unit 2.
5 is output. In the visualization unit 36, for example, each group is represented by an ellipse, and the semantic distance between each group is
Information such as the number of included information items, the number of included related words, and the semantic spread of included related words can be added and displayed. For example, some or all of the above information,
It can be visualized by the size and color of the ellipse, the depth of color, the distance between the ellipses, and the arrangement.

FIG. 15 is an explanatory diagram of an example of a display result in the fifth embodiment of the present invention. FIG. 15 shows an example of the display result on the output unit when the initial set is selected using the word “PDA” as a key and the “means” is classified as the viewpoint as in the above example. Here, as shown in FIG. 14, since “management”, “communication”, and “client server system” exist in the same hierarchical level, these are represented by ellipses, and the number of information included is the size of the ellipse. Is expressed in. Also, each word is added to each ellipse.

In this way, when the initial set is classified and the initial search space is visualized, the user can perform various operations for information search. As the search operation, for example, “fine classification with the same viewpoint”, “fine classification with a viewpoint added”,
"Change attention word", "Change viewpoint", "History operation",
"Reduction of the basic part" is possible. The possible operations are displayed as a menu on the output unit 25, for example, and the user can execute the search operation by selecting the displayed menu with, for example, a mouse.

First, the operation of the search mechanism when the user selects "fine classification from the same viewpoint" will be described. When the user selects “subclassification from the same viewpoint” from the menu, the search mechanism 23 prompts the user to select one of the visualized groups according to a message displayed on the output unit 25 or the like. Urge.

When a group is selected, the selected space is designated as the “selected group” by the search space holding unit 32.
The hierarchical structure group that is held in the group and is included in the selected group is sent to the grouping unit 35, and is grouped in the same manner as in the initial set. The grouped result is sent to the visualization unit 36, visualized and displayed on the output unit 25.

FIG. 16 is an explanatory diagram showing an example of a display result at the time of subclassification according to the fifth embodiment of the present invention. FIG.
When the "communication" group is selected in the state where the visualization display shown in is displayed, the groups are grouped in the lower hierarchy. In the case of the hierarchical structure shown in FIG. 14,
A group of "FAX", "electronic mail", and "personal computer communication" under "communication" is formed and visualized as shown in FIG. In this way, the user can display the subdivision of the visually displayed group, and can narrow down the search from conceptual words to more specific words.

When the user specifies a group and then selects "fine classification from the same viewpoint", the user is not prompted to select the group, and the grouping is performed. Alternatively, since “sub-classification from the same viewpoint” is an operation that may be executed frequently, it may be configured to be executed by directly selecting a group without selecting a menu.

Next, the operation in the case of selecting "fine classification with viewpoint added" will be described. When the user selects “sub-classification with additional viewpoint” from the menu, the user is prompted to select a group with a message or the like, as in the case of “sub-classification with the same viewpoint”. If a group has already been selected, that group will be the selected group. When a group is selected, viewpoints other than those used in the current classification are displayed, prompting the user to select a viewpoint to add. When the user selects a viewpoint, the hierarchizing unit 34 hierarchizes the selected group in the same manner as when classifying the initial set, the grouping unit 35 groups the hierarchies, and the visualization unit 3
Visualize at 6. In this way, the classification results are displayed after being classified by the added viewpoint. By this processing, the group narrowed down from the viewpoint used when classifying the initial set can be classified from another viewpoint, and the information can be effectively narrowed down.

Next, the operation when the user selects "change noticed word" will be described. When the user selects "change noticed word" from the menu, the user is prompted to select a group with a message or the like, as in the case of "fine classification from the same viewpoint".
If a group has already been selected, that group will be the selected group. When the group is selected, the user is prompted to select the attention word. The user specifies the attention word to be changed. For example, in the state where the display shown in FIG. 16 is made, the central concept of the selected group is “communicate with PDA” and the attention word is “PDA”. Change the word of interest. Then, the viewpoint extraction unit 33 extracts the relationship that the new attention word “communication” has with other words, creates a viewpoint list, and presents it to the user. The user selects a viewpoint to be classified from the viewpoints, and based on the selected viewpoint, hierarchization by the hierarchizing unit 34 is performed in the same manner as when the initial set is classified,
Grouping by the grouping unit 35 and visualization by the visualization unit 36 are performed. In this way, it is possible to perform classification by different attention words within the same set.

Next, the operation when "change viewpoint" is selected will be described. User can change the viewpoint from the menu
When you select, the selectable viewpoint is displayed. When the user selects a viewpoint from these, the current classification is discarded and classification is performed with the newly selected viewpoint. This allows
Different viewpoints can be classified within the same set.

Next, the operation when the "operation of history" is selected will be described. When a user performs a search operation, information necessary for the operation and return is accumulated as a history. When the user selects "history operation" from the menu, the state of the previous screen is restored based on the accumulated history. As a result, the user can return to the state before the operation.

Next, the operation when "reduction of the basic part" is selected will be described. In the “reduction of the basic part”, if there is a word that the user does not need among the words expanded when the initial set is selected, information including the word is excluded. When the user selects "reduction of basic part" from the menu, the key word entered and the word expanded by the thesaurus are displayed. By selecting and instructing an unnecessary word, the user excludes information including the instructed word from the search space. For example, if the word "computer" is expanded to the words "computer", "workstation", "personal computer", etc., select "Reduce basic part" to display a list of these words. . If the information about the "personal computer" is unnecessary, the information including the "personal computer" can be deleted from the search space by designating this without changing the overall classification.

These search processes are not only performed for each initial set, but further search processes can be performed in a state after performing a certain process. For example, "change attention word" or "change viewpoint"
After performing the above, it is possible to perform “sub-classification from the same viewpoint” or “sub-classification with additional viewpoint”. In addition, it is possible to instruct the sub-classification to the group after performing the “sub-classification from the same viewpoint” and the “sub-classification with addition of the viewpoint”. At this time, if the lowest group in the hierarchical structure, that is, one piece of text information is instructed, further subclassification cannot be performed. Therefore, a message indicating that the subclassification is impossible is output, or the text is displayed. It may be configured to display information.

By appropriately performing such a search process and performing classification using various viewpoints and various attention words, information is narrowed down while grasping the whole even if the search intention is not always clear. You can easily get the information you want.

The process of narrowing down the information in the fifth embodiment of the present invention will be described below using a specific example. FIG.
FIG. 7 is an explanatory diagram of a specific display example of search results according to the fifth embodiment of this invention. Here, it is assumed that “automobile” is input as the attention word and classification is performed according to the operation target of “automobile”. That is, the classification is made as "what to do with the automobile". In FIG. 17, each classified group is indicated by an ellipse having a size according to the number of pieces of information belonging to the group. The display form of the group is not limited to the ellipse, and any display form may be used as long as the display shows at least the existence of the group.

In FIG. 17, the column of noticed words at the top is an input region for noticed words. Each button in the second row indicates an operation mainly based on the target word, an operation targeted, an operation defined as a means, an intended operation, an operation location, and an operation attribute. Is selected. Further, each button in the third row indicates the subject, target, means, purpose, place, and attribute of the action of the attention word, and the button indicated by a broken line indicates that there is no corresponding viewpoint. Note that the buttons indicated by the broken lines cannot be selected. For example, it indicates that there is no object to perform the operation of “car”, that is, there is no relationship of “caring”.

FIG. 18 is an explanatory diagram of a display example when changing the viewpoint of a specific search result in the fifth embodiment of the invention. In the display example shown in FIG. 17, for example, when the button for selecting the attribute of the word of interest in the third row from the top is pressed with, for example, the mouse, the viewpoint changes to the attribute of the word of interest, that is, the classification according to the relationship of “automobile”. To be done. As a result, the classification result as shown in FIG. 18, for example, is displayed.
In FIG. 18, since the number of groups in the same hierarchy is small, a group of words expanded below “component” is also displayed. In this way, the viewpoint can be easily changed by pressing the button on the upper part of the display screen.

FIG. 19 is an explanatory diagram of a display example of a specific search result at the time of fine classification in the fifth embodiment of the present invention. On the other hand, in the display example shown in FIG. 17, by selecting a group, the selected group can be subdivided. For example, by selecting the group “sale” from the groups displayed in FIG. 17, the classification result as shown in FIG. 19 can be obtained. Here, it is assumed that the subject of the attention word is selected as a default as a viewpoint when selecting a group. In other words, they are classified from the perspective of "who sells the car?" As shown in FIG. 19, groups of "makers", "dealers", and companies that are subordinate to them are displayed as ellipses as the main bodies that sell automobiles.

On the display screen shown in FIG. 19, the selectable viewpoints are larger than in the display example shown in FIG. In this example, “selling a car” can be the subject of something. For example, "the sales of automobiles will increase" and "the sales of automobiles will be promoted" can be considered. Therefore, some of the buttons on the second row can be selected. 3
The button in the second row is in a selectable state in order to select a subject or an object for “selling a car”.

FIG. 20 is an explanatory diagram of a display example at the time of fine classification in which the viewpoint of a specific search result is changed in the fifth embodiment of the present invention. In the display example shown in FIG. 19, by pressing the upper button, the viewpoint of subclassification can be changed. For example, by pressing a button indicating the location of the word of interest in the third row, it is possible to classify from the viewpoint of "where to sell the car". The result of this classification is shown in FIG. Results are shown categorized by where the car is sold.

By selecting a group in the state shown in FIG. 20, further detailed classification can be performed. For example, by selecting the group “Japan”, it is possible to obtain a set of information described regarding “Sell a car in Japan”, that is, “Sell a car in Japan”.

If the number of information groups becomes sufficiently small by performing such fine classification and change of viewpoints, it becomes unnecessary to classify further viewpoints. FIG. 21 is an explanatory diagram of a display example when a specific search result has a small number of classified information in the fifth embodiment of the present invention. As a result of the classification, when the number of included information is small, instead of displaying by group as in the above example, a list of information is presented together with the target sentence as shown in FIG. 21, for example. Can be displayed. And
By selecting any one from this list, the contents of the information body can be referred to. The number of information items to be displayed in the list format can be set by the user. The default value can be set to about 100, for example.

FIG. 22 is an explanatory diagram of another example of the configuration of the display screen in the fifth embodiment of the invention. Although the example of displaying the classification result is shown in the above example, various areas can be arranged on the display screen in addition to the selection area indicating the classification result. In the example shown in FIG. 22, in addition to the group selection area for displaying the classification result as in the above example, a next screen area in which the selected group is sub-classified is provided, and the sub-classification result is displayed together with the original classification result. Shown at the same time. This allows easy selection. Also, here, the history of operations such as fine expansion and change of viewpoint is shown as small icons. By clicking on any of these, you can freely move to the past state. Further, in this example, an information content display area is provided, and while referring to the content of information, it is possible to refer to the selection state of the group that has reached the time when the information is selected. This information content display area is not necessary until the content of information can be selected, so for example, when narrowing down a group, the group selection area and the next screen area are largely arranged and narrowed down without providing this area. So that you can easily perform operations,
You may make it change the structure of a display screen suitably. Of course, the screen configuration is not limited to this, and the screen may be configured to be easy to use.

FIG. 23 is a block diagram showing the sixth embodiment of the information search system of the invention. In the figure, the same parts as those in FIG. 10 are denoted by the same reference numerals, and description thereof will be omitted. Reference numeral 37 is a dependency extraction unit. In the above-described fifth embodiment, the text information stored in the information storage unit 21 is analyzed in advance, the analysis result is stored in the analysis result storage unit 22, and a search is performed using this analysis result. . The sixth embodiment shows an example in which text information is analyzed when performing a search.

As in the fifth embodiment, the initial set selection unit 31 expands a key word when it is input, using a thesaurus. Furthermore, the initial set selection unit 21
The information storage unit 21 is searched for sentences including these words.
As a search method, a method such as a full text search method or an index method may be used.

The dependency extraction section 37 analyzes a part of the text of each piece of information obtained as a result of the retrieval, and extracts the dependency of the word used in the retrieval. The part of the text is the range of dependency analysis centered on the search key,
Analyze one sentence, two sentences, one paragraph, etc. within a predetermined range. Of course, the entire text may be parsed if desired. For analysis, various conventionally used analysis methods can be used.

The viewpoint extracting unit 73 and the hierarchizing unit 74 perform the viewpoint extraction and the related word extraction based on the analysis result.

The functions and the overall operation of the other units are the same as those in the above-mentioned fifth embodiment, and therefore will be omitted here. According to the sixth embodiment, the analysis result storage unit 2
Since 2 is not required, the storage capacity can be reduced, and the analysis amount can be reduced because it is not necessary to analyze the entire text. However, the time required to form the search space in the initial set selection unit 21 is shorter in the configuration described in the fifth embodiment.

Further, by using both the configurations of the fifth and sixth embodiments described above, a light analysis such as a morphological analysis is performed in advance and stored in the analysis result storage unit 22, which is further required at the time of retrieval or classification. Degree extractor 37
You may do it in. Further, instead of referring to the entire text, a plurality of phrases or sentences representing the subject of the text may be extracted in advance, and this may be used as a target for searching and classification. In addition, the subject of the extracted text is subjected to dependency analysis and a subject index is added, and the subject index is used to search for text that satisfies the search request. Also, the subject index part has a specified relationship. You may comprise so that a phrase may be extracted.

[0135]

As is apparent from the above description, according to the present invention, a search is performed from the viewpoint desired by the user based on the meaning and context of the expression used for the search in the text. The results can be categorized and displayed, and the user can search for a desired information from among a huge amount of information. At this time, the user can be configured to narrow down while classifying by changing the viewpoint and the attention word, and the user can search for desired information while grasping the whole. effective.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a first embodiment of an information search device of the present invention.

FIG. 2 is an explanatory diagram of an example of an operation and an intermediate process when “release” is input as a search request in the first embodiment of the present invention.

FIG. 3 is an explanatory diagram illustrating an example of a display of classification results according to the first embodiment of this invention.

FIG. 4 is a configuration diagram showing a second embodiment of the information search device of the present invention.

FIG. 5 is an explanatory diagram showing an example of an operation and an intermediate process when “release a workstation” is input as a search request in the second embodiment of the present invention.

FIG. 6 is a configuration diagram showing a third embodiment of the information search device of the invention.

FIG. 7 is an explanatory diagram of an example of a thesaurus.

FIG. 8 is an explanatory diagram illustrating an example of a display of a classification result according to the third embodiment of the present invention.

FIG. 9 is a configuration diagram showing a fourth embodiment of the information search device of the invention.

FIG. 10 is a configuration diagram showing a fifth embodiment of the information search device of the invention.

FIG. 11 is an explanatory diagram of an example of an analysis result.

FIG. 12 is an explanatory diagram of the outline of an example of an operation in the fifth embodiment of the information search system of the invention.

FIG. 13 is an explanatory diagram of an example of a viewpoint conversion table.

FIG. 14 is an explanatory diagram of an example of layering based on a thesaurus.

FIG. 15 is an explanatory diagram showing an example of a display result according to the fifth embodiment of the present invention.

FIG. 16 is an explanatory diagram showing an example of a display result at the time of sub-classification according to the fifth embodiment of the present invention.

FIG. 17 is an explanatory diagram of a specific display example of search results according to the fifth embodiment of the present invention.

FIG. 18 is an explanatory diagram of a display example when changing the viewpoint of a specific search result according to the fifth embodiment of the present invention.

FIG. 19 is an explanatory diagram of a display example when a detailed search result is finely classified according to the fifth embodiment of the present invention.

FIG. 20 is an explanatory diagram of a display example when fine classification is performed by changing the viewpoint of a specific search result according to the fifth embodiment of this invention.

FIG. 21 is an explanatory diagram of a display example when the number of classified information items of a specific search result is small in the fifth embodiment of the invention.

FIG. 22 is an explanatory diagram of another example of the configuration of the display screen according to the fifth embodiment of the present invention.

FIG. 23 is a configuration diagram showing a sixth embodiment of the information search device of the invention.

[Explanation of symbols]

1 ... Information storage unit, 2 ... Search request input unit, 3, 11 ... Search unit, 4, 12 ... Dependency extraction unit, 5 ... Storage unit, 6 ... Output unit, 7, 14 ... Viewpoint input unit, 8, 15 , 18 ... Information classification unit, 10 ... Information retrieval unit, 13 ... Relationship comparison unit, 16 ... Thesaurus, 17 ... Classification selection unit, 21 ... Information storage unit, 22 ...
Analysis result storage unit, 23 ... Search mechanism, 24 ... Input unit, 25
... Output unit, 26 ... Thesaurus holding unit, 31 ... Initial set selecting unit, 32 ... Search space holding unit, 33 ... Viewpoint extracting unit, 34
... hierarchization unit, 35 ... grouping unit, 36 ... visualization unit, 3
7 ... Dependency extraction unit.

Claims (13)

[Claims] 1. An information storage means for storing information including at least text information, a search request input means for inputting a search request, and an information search means for searching for information including the search request in the text information. A viewpoint input means for inputting a relationship that the phrase of the search request may have with other phrases in the text information as a viewpoint for classifying the information retrieved by the information retrieval means; In the text information of each information, an information classification unit that extracts a phrase having the relationship with the phrase of the search request and classifies the retrieved information based on the difference between the extracted phrases, and the information retrieval unit retrieves the information. And a storage unit that stores the classified information obtained by the information classifying unit and an output unit that outputs the searched information and the classified result. Information search device to collect. 2. An information storage unit for storing information including at least text information, a search request input unit for inputting a phrase as a search request, and the phrase input as a search request by the search request input unit. Information retrieval means for retrieving information contained in text information, viewpoint input means for inputting a relationship that the word and phrase may have with other words and phrases in the text information as a viewpoint for classifying the information, and the information retrieving means The information in the information group based on the difference between the extracted words and phrases in the text information of each information of the information group forming all or part of the information retrieved by Information classifying means for classifying, and a classification result selecting means for selecting one or more groups from the results classified by the information classifying means,
And a storage unit for storing the information retrieved by the information retrieval unit, the classification result by the information classification unit and the selection result by the classification result selection unit, and an output unit for outputting the retrieved information and the classification result. The information belonging to one or a plurality of groups selected by the classification result selecting means is used as an information group, and the information classifying means can perform classification based on the relationship input by the viewpoint inputting means. Information search device. 3. An information storage means for storing information including at least text information, a search request input means for inputting two expressions as a search request and an expression representing a relationship between them, and an expression satisfying the search request. And an information retrieval means for retrieving the information having in the text information, and one of the two terms as a viewpoint for classifying the information retrieved by the information retrieval means, and the designated term is A viewpoint input means for specifying a relationship different from that used in the search among other words and phrases, and a word or phrase specified by the viewpoint input means in the text information of each information searched by the information search means And an information classifying unit that classifies the retrieved information based on a difference between the extracted phrases having a relationship specified by the viewpoint inputting unit, and the information. An information search apparatus comprising: a storage unit that stores information searched by a search unit and a classification result by the information classification unit; and an output unit that outputs the searched information and the classification result. 4. An information storage means for storing information including at least text information, a search request input means for inputting two expressions as a search request and an expression indicating a relation between them, and an expression satisfying the search request. And an information retrieval means for retrieving the information having in the text information, and a relation in which one of the two phrases is designated as a viewpoint for classifying the information and the designated phrase may have another phrase. Of the phrase extracted by extracting the phrase having the relation with the phrase in the text information of each information of the information group forming all or a part of the information retrieved by the information retrieving device. An information classifying unit that classifies the information in the information group based on a difference, and one or a plurality of groups from the classification results classified by the information classifying unit. Classification result selecting means for selecting, storage means for storing the information searched by the information searching means, the classification result by the information classifying means and the selection result by the classification result selecting means, the searched information and the classification result The information classifying means based on the relationship input by the viewpoint inputting means, the information belonging to one or more groups selected by the classification result selecting means as an information group. An information search device characterized by enabling classification. 5. An analysis result that stores a result of analyzing a relationship between an information storage unit that stores at least information including text information and a phrase of the text information stored in the information storage unit. A storage space, a thesaurus holding unit that holds a thesaurus showing the hierarchical relationship between words, and text information including the input clue words are searched based on the analysis result in the analysis result storage unit to search the search space. An initial set selection unit that is selected as information to be configured, a search space holding unit that holds the text information searched by the initial set selection unit in a searchable manner, and a relationship that the clue phrase can have with other phrases in the text information. Is stored in the thesaurus holding unit, and a point extraction means for extracting a plurality of points is stored in the thesaurus holding unit. An information search device characterized by having a hierarchizing means for hierarchizing based on a thesaurus, a grouping means for grouping into one or more groups based on a hierarchical structure, and a visualizing means for visualizing groups. . 6. An information storage unit for storing information including at least text information, a thesaurus holding unit for holding a thesaurus showing a hierarchical relation of meanings of words, and text information including an input clue word and phrase. An initial set selection unit that searches from the information storage unit and selects it as information that forms a search space; and a search space holding unit that holds the text information searched by the initial set selection unit in a searchable manner,
Dependency extracting means for analyzing the text information including the clue word and extracting the dependency relationship of the clue word and the clue word in the text information based on the dependency relationship extracted by the dependency extracting means At the viewpoint extraction means for extracting a plurality of possible relationships with other words, and based on the thesaurus held in the thesaurus holding unit words having a relationship corresponding to the viewpoint in which the clue word is input in the text information An information searching apparatus comprising: a hierarchizing means for hierarchizing, a grouping means for grouping into one or more groups based on a hierarchical structure, and a visualizing means for visualizing the groups. 7. The information search device according to claim 5, wherein the clue word is a word developed by using a thesaurus held in the thesaurus holding unit, which is a word entered by a user. . 8. The information searching apparatus according to claim 7, wherein the clue word is only a part of the expanded word according to a user's instruction. 9. The information search device according to claim 5, wherein the grouping unit groups one or more groups instructed by a user in a lower hierarchy and visualizes the group by the visualization unit. . 10. The information searching apparatus according to claim 5, wherein the hierarchizing unit performs hierarchization based on the new viewpoint when the user inputs a new viewpoint. 11. The hierarchizing means, when the user designates one or more groups and inputs a new viewpoint, hierarchizes the designated groups based on the new viewpoint. The information search device according to claim 5. 12. The viewpoint extracting means extracts, when the user specifies a new clue word, a plurality of relationships that the new clue word can have with other words in the text information, and the hierarchizing means The information according to claim 5, wherein the new clue word is hierarchized based on the thesaurus held in the thesaurus holding unit, with the word having a relation corresponding to the viewpoint input in the text information. Search device. 13. The dependency extraction means, when the user specifies a new clue word, analyzes the text information including the new clue word, extracts the dependency relationship of the new clue word, and The viewpoint extracting means extracts a plurality of relationships that the new clue word / phrase may have with other words / phrases in the text information based on the dependency relationship extracted by the dependency extraction means, and the hierarchizing means is The information search according to claim 6, wherein a phrase having a relationship corresponding to a viewpoint in which a new clue phrase is input in the text information is hierarchized based on the thesaurus held in the thesaurus holding unit. apparatus.