JP2009217436A - Collaborative sorting apparatus and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To increase the rate of reproducing a search process where a sorting axis is designated by facilitating the sorting of sorting axes that are likely to diverge, in a collaborative sorting system. <P>SOLUTION: Processes for conceptually associating input text data while referring to previously stored sorting axes and the content to be sorted, then indexing and grouping the data, storing the data in a storage means, and sending and presenting a group index or a hierarchy index to a user from the storage means according to a bundle name inputted from the user are repeated until the user inputs an instruction to not continue hierarchization. When a search query is inputted from the user, a normal search is conducted based on the search query, the storage means is referenced based on the search query, and the search query is expanded for the search. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a collaborative classification apparatus and program, and more particularly to a collaborative classification apparatus and program in which a plurality of users classify and share information such as bookmarks, photos, videos, books, and papers.

  Recently, collaborative tagging systems (Collaborative Tagging Systems) that enable collection of highly organized information by sharing the results of each user organizing and classifying information such as URL (bookmark) photos, videos, and papers. Is prosperous. In such a system, users can freely assign tags (classification axes), so it is possible to naturally guide users to the information sharing system from the act of using the Internet, and participate in the service. Because the threshold is low, it attracts many users.

  Tag classification is called Folksonomy, and its positioning has been discussed in various ways, and numerous results have been published in research.

  In an existing cooperative classification system, a tag list used by a user is often displayed in a tag ground format. Tags input by the user are free, but tend to explode in kind, and tags that are simply arranged in the order of Japanese syllabary or frequency do not function effectively as a classification axis when searching for information. If only frequent items are listed, it will be difficult to find detailed tags.

To solve these problems, social bookmarking services “del.icio.us” and “goo bookmark” provide a tag grouping function. In these services, an interface for grouping tags is provided, and tags can be grouped as desired by the user himself (for example, see Non-Patent Document 1 and Non-Patent Document 2).
http://del.icio.us http://bookmark.goo.ne.jp

  However, in such a conventional technique, since there are so many types of tags, the task of grouping / hierarchizing the tag list is not easy. The user must determine which tags are close to each other by matching each other.

  In addition, the user can search the collaborative classification system using the tag as a search query. However, when the frequency of use of the tag selected by the user is low, the obtained information decreases, and the search reproduction rate. There is a problem that goes down.

  The present invention has been made in view of the above points, and is used when performing a search by tag to determine what kind of grouping should be performed in a technology bundiling function for collecting and organizing divergent tags. Resolved the problem that search was performed using infrequent tags, resulting in too few results, facilitated the organization of classification axes that tend to diverge in a collaborative classification system, and reproduction of search processing with a specified classification axis An object of the present invention is to provide a collaborative classification apparatus and program capable of increasing the rate.

  FIG. 1 is a principle configuration diagram of the present invention.

The present invention (Claim 1) is a collaborative classification information processing apparatus for a plurality of users to classify and provide information such as bookmarks, photographs, videos, books, papers, etc.
User interface means 10 as a data input / output unit;
Communication means 80 for performing data communication;
A database 65 storing a classification axis including tags, images and sounds, classification targets, and user information classified by a plurality of users;
When text data is input from the user's client device, the contents of the database 65 are referred to, the text data is conceptually related, and association processing means 20 for storing in the index database 15;
Indexing processing means 30 for indexing and grouping the associated data in the index database 15 and storing the data in the index database 15.

Further, the present invention (Claim 2) is provided in the association processing means 20,
An element extraction processing means for extracting co-occurrence data having a characteristic property from text data with reference to the database 65;
Filtering means for removing noise using a predetermined minimum frequency condition of elements in the co-occurrence data;
Vector calculating means for obtaining a probability vector of each element of the co-occurrence data from which noise has been removed.

Further, the present invention (Claim 3) is provided in the indexing processing means 30.
Neighborhood index assigning means for indexing an element close to an element designated by the user as a neighborhood index;
Group index assigning means for indexing elements belonging to the same group as the element designated by the user as a group index;
And a hierarchical index assigning means for indexing a lower element group and an upper element of the element designated by the user as a hierarchical index.

  Further, according to the present invention (Claim 4), the group index or the hierarchical index of the index database 15 is transmitted to the user's client device based on the bundle name input from the user, and the process of presenting the hierarchical index from the user is hierarchically And bundling / hierarchical recommendation means that repeats until an instruction not to continue is input.

Further, according to the present invention (Claim 5), when a search query is input from a user, a normal search is performed based on the search query, and the index database 15 is referred to based on the search query. Search query expansion means 50 for expanding and searching a search query;
It further has a result switching display means 90 for switching and displaying the search result using the normal search result and the extended search query based on a user instruction.

  The present invention (Claim 6) is a cooperative classification information processing selection program for causing a computer to function as each means constituting the cooperative classification information processing apparatus according to any one of claims 1 to 4.

  The present invention facilitates the organization of classification axes that tend to diverge in a cooperative classification system, and can increase the recall of a search process that designates a classification axis.

  First, terms used in this specification will be described.

-Collaborative classification system: A system that allows end users to freely classify information and share classification results. Example: Social bookmarking service (del.icio.us, Hatena, etc.,), photo (flickr), video, paper (citeulike), etc.,
SBM: Abbreviation for social bookmark service. The classification target is a URL. A system for sharing bookmarks of each user over a network. A typical example of a collaborative classification system.

  Data model of SBM: In this embodiment, a three-set model in which (USER), which URL (RESOURCE), and what category (TAG) are classified. Assume UxRxT.

  Classification axis: Information used for classification in a cooperative classification system. Tags (keywords), images, and sounds are used as the classification axis.

  -Classification target identifier: The classification target needs to be distinguishable for classification. Being able to distinguish between classification targets is synonymous with distinguishing classification targets by identifiers because they can have a one-to-one correspondence with identifier sets that can be uniquely determined. In the case of a bookmark, the URL can be used as an identifier. That is, in the present embodiment, no particular problem occurs even if the classification target is represented by an identifier.

  Tag: Classification axis in SBM. An arbitrary word can be determined by each user.

  -Tag bundling: Link a tag with a certain keyword (bundle name) (see del.icio.us (http://del.icio.us)). A lower / upper / lower relationship is not always required between a bundle name and a tag. In many cases, multiple attribution is also permitted. One way to organize tags.

  -Tag hierarchization: Hierarchical arrangement of tags or recursive bundling (see youtube (http://youtube.com)).

  Query: In this embodiment, a search tag for a given tag is used.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  The basic operation of the present invention will be described.

  As shown in FIG. 2, the cooperative classification system of the present invention classifies a classification target (URL) by a user with a certain classification axis (tag).

  The present invention is a result of classification by a user in a cooperative classification system, and is appropriate when organizing the classification axis by conceptually indexing related information of the user / classification axis / classification target in advance. Expected groupings and hierarchies are recommended, and search queries are expanded based on concepts during search. This recommendation is indexed in advance and information can be acquired at high speed. Thereby, the said subject can be solved.

  The basic operation of the present invention is as follows.

  (1) Using the association (grouping hierarchization) between the classification axes given in advance, when the user organizes the classification axes, the group and the hierarchical structure are recommended.

  (2) For the search query for the classification axis, the search result is expanded using the relevance of the classification target to the classification axis given in advance.

  FIG. 3 is a schematic block diagram of a system according to an embodiment of the present invention.

  The cooperative classification system is provided to the client device 2 as a network service, and the user can use the service through a Web browser or a client application. In a cooperative classification system (information sharing server), an application server 10 that actually performs processing, database servers 15 and 60 that store data, an association engine 30 that associates classification axes and classification targets, user information, and association information as an application The indexing engine 30 and the query expansion engine 50 are pre-indexed according to the internal logic.

  These processing units are not only realized in a single server, but may be distributed in a plurality of units.

  Hereinafter, the information sharing server 1 shown in FIG. 3 will be described as a cooperative classification device.

  FIG. 4 shows the configuration of the cooperative classification apparatus according to an embodiment of the present invention.

  The cooperative classification apparatus shown in the figure includes a communication interface (I / F) 10, an association processing unit 20, an indexing processing unit 30, a bundling / hierarchical recommendation processing unit 40, a query expansion unit 50, a classification axis DB 60, a classification. It consists of a target DB 70, a user DB 75, a communication unit 80, a search result display control unit 90, and an index DB 15.

  The association processing unit 20 uses all or any of the classification target / tag (classification axis) / user information of the classification axis tag 60, the classification target DB 70, and the user DB 75, and extracts characteristic features from them, for example, It is possible to measure distances such as a tag tag and a classification target tag. For this purpose, Reference 1 “Probabilistic Latent Semantic Analysis”, Thomas Hofmann, 1999, In Proc. Of Uncertainty in Artificial Intelligence, UAI'99, Reference 2 “Exploring Social Annotations for the Semantic Web”, Xian Wu et. al, 2006, In WWW2006 ”, Reference 3“ Visualization of SBM user-tag relations using PLSI, Takashi Mao, Yasuharu Eda, Masatoshi Yoshikawa, Masashi Yamamuro, DBWS, 2007 ” can do. Typically, each item is expressed as a feature vector, and the distance between elements is measured using the similarity between the vectors.

  As shown in FIG. 5, the association processing unit 20 includes an element extraction unit 21, a pre-filter processing unit 22, and a PLSI (Probabilistic Latent Semantic Indexing) processing unit 23, and the result of each processing unit is stored in the index DB 15. And

  FIG. 6 is a flowchart of the associating process according to the embodiment of the present invention.

  First, when raw text data is input via the communication interface 10, the association processing unit 30 extracts necessary co-occurrence data from the raw data in the element extraction processing unit 21 (step 101), and performs pre-filter processing. The unit 22 removes noise data from the extracted co-occurrence data (step 102). The PLSI processing unit 23 performs indexing on the co-occurrence data from which noise has been removed (step 103), and as a result, outputs the feature vector of each element to the index DB 15 (step 104).

  Hereinafter, detailed processing of the association processing unit 20 will be described.

  Raw text data is input to the element extraction unit 21. The raw data structure in social bookmarks depends on how the database schema is defined within each service. As a specific example, the act of bookmarking is classified as “who (U), which URL (R), when (t), what” (T). (C) ”, and the raw data can be regarded as five sets of co-occurrence data (subsets of U × R × t × T × C) each consisting of a direct product. . In the method of the above-mentioned document 1, a stochastic indexing method composed of two sets is adopted, and five sets of co-occurrence data are selected so that the indexing method can be executed (such as U × R and R × T). . Also, in the above-mentioned literature 2, literature 3, and literature 4, “Toward the construction of an automatic classification system using the tags of Folksonomy”, Masaharu Eda, Masatoshi Yoshikawa, Masashi Yamamuro, DBWS, 2007, three sets co-occurrence Since indexing can be performed on the data, three sets of co-occurrence data such as (U × R × T: user, classification axis, classification target) are selected. These processes are processes performed by the element extraction processing unit 21, and information not selected is simply ignored. The element extraction processing unit 21 stores the selected co-occurrence data in the co-occurrence data area of the index DB 15.

  The PLSI, which is the indexing technique of the above-mentioned literature 1, is expanded from 2 sets of co-occurrence data to 3 sets. This is the method of the above-mentioned literature 2, and the discussion is similarly established. I will explain.

  Next, the pre-filter processing unit 22 will be described.

  In PLSI, it is known that items with low appearance frequency in co-occurrence data become noise (see Document 2). Therefore, the pre-filter processing unit 32 reads the co-occurrence data selected by the element extraction unit 21 from the index DB 15 and performs noise removal from the co-occurrence data using the minimum frequency condition of the elements in the co-occurrence data. And stored in the index DB 15. Here, each item frequency will be described.

Now, specific examples of three sets of data {(u1, r1, t1), (u1, r1, t2), (u1, r2, t1), (u2, r2, t2), (u3, r2, t3) }
And In this case, each item frequency represents the number of times an item appears in the data set as three sets. That means
│u1│ = 3, │u2│ = 1, │u3│ = 1
│r1│ = 2, │r2│ = 3
│t1│ = 2, │t2│ = 2, │t3│ = 1
It becomes. However, | x | represents the density (number) of the item x. In this case, if “2” is specified as the minimum frequency, u2, u3, t3 are removed as noise, and the three sets of data are {(u1, r1, t1), (u1, r1, t2)}.
Remains.

  Next, the PLSI processing unit 23 will be described.

  The PLSI processing unit 33 reads the co-occurrence data from which noise has been removed by the pre-filter processing unit 22 from the index DB 15, and uses the PLSI proposed in the above-mentioned literature 1 and literature 2 to generate N sets of co-occurrence data. The co-occurrence is learned, the item of each element is obtained as a probability vector, and stored in the index DB 15. Refer to Documents 1 and 2 for details of the processing. Here, the probability vector is a normalized vector such that the sum of the values of a certain vector is 1. By using KL divergence or JS divergence as the distance between vectors, it is possible to measure the distance between items with high accuracy.

  Next, the indexing processing unit 30 will be described with reference to FIG.

  As shown in FIG. 5, the indexing processing unit 30 includes a neighborhood indexing unit 31 and a grouping hierarchizing unit 32, and each result is stored in the index DB 15.

  The indexing processing unit 30 performs the following three patterns.

  1. The neighborhood indexing unit 31 reads the feature vector (PLSI vector) obtained by the association processing unit 20 from the index DB 15, indexes an element having a distance close to the element designated by the user, and uses this as a neighborhood information index. Is stored in the neighborhood index area.

  2. The grouping / hierarchizing unit 32 reads a feature vector (PLSI vector) from the index DB 15, indexes elements belonging to the same group as the element designated by the user, and stores them as a group index in the group index area of the index DB 15. .

  3. The grouping / hierarchizing unit 32 reads the feature vector (PLSI vector) from the index DB 15, indexes the lower element group of the element designated by the user, and stores it as a hierarchical index in the hierarchical index area of the index DB 15.

  The details are performed using the methods shown in the above-mentioned documents 3 and 4.

  An example of an index that can be obtained as a result is shown below.

(1) "web2.0" --- "web", "Internet", "internet", "www", "html", "blog", "SBM"
(2) "web2.0" --- 5
“5” above indicates a group ID. The group ID is a number assigned to each cluster by clustering tags based on the feature amount. As tags belonging to the number “5”, {“web”, “Internet”, “internet”, “www”} and the like can be assumed.

(3) "web2.0" --- Parent->, Children-> (blog, thml, SBM ,,,)
-For the route, the top tag candidate is returned as a child. (The method for configuring the top tag candidate is described in Document 4)
* "" --- parent->"",child-> (web, read later, tools, reference ,,,)
Inquiry processing is speeded up by calculating such an index in advance in the indexing processing unit 30.

  Next, the bundling / hierarchical recommendation processing unit 40 will be described.

  First, the bundle recommendation process will be described. The bundling / hierarchical recommendation processing unit 40 typically uses a group index at the time of bundling and a hierarchical index at the time of hierarchization.

  FIG. 7 is a flowchart of the bundling recommendation process according to the embodiment of the present invention.

  First, the bundle name name and the entire tag set T are input from the user client device 2 via the user interface 10 (step 201). In the bundle recommendation process, the bundle name name may be a tag that has already been used or may be an arbitrary keyword. In the case of a tag (step 202, Y), a list of tags included in the same group is presented to the user as a bundle target candidate using the group index stored in the index DB 15 (step 205). In the case of an arbitrary keyword (step 202, N), by letting the user select one tag (steps 203, 204), a set Rt of bundle candidate tags is obtained using the group index of the selected tag. (Step 205) and present it to the user (step 206). When the above group index is used, the same recommendation can be made even if it is replaced with a neighborhood information index or a hierarchical index. In the case of a hierarchical index, it is possible to more intuitively recommend bundles by acquiring information using the bundle name as a parent tag.

  FIG. 8 shows the flow of the user interface at the time of bundling in an embodiment of the present invention. The flow of the user interface shown in the figure is a bundling procedure when a group index is used.

  (1) The user displays a bundle creation menu on the client device 2 and inputs a new bundle name.

  (2) Next, referring to the memory 33 based on the input bundle name, a bundle recommendation tag for inputting a tag included in the bundle is displayed.

  (3) The user selects the bundle recommendation tag displayed or inputs an arbitrary tag by manual input.

  (4) A tag is input manually.

  Next, the hierarchical recommendation process of the bundled / hierarchical comment processing unit 40 will be described. FIG. 9 is a flowchart of the hierarchical recommendation process according to the embodiment of the present invention.

  Since the hierarchization operation is usually a recursive repetitive operation, the hierarchization recommendation process in FIG. 9 can be recursively hierarchized by selecting whether or not to continue the hierarchization process. Yes.

  First, a parent candidate (empty when a top tag is selected) p is input (step 301).

  The lower tag set Rt is acquired from the index DB 15 using p as a key (step 302), and the set Rt is presented to the user. At this time, a selection form capable of selecting an arbitrary tag is presented (step 303). When continuing the hierarchization process by the user (step 304, Y), the parent candidate selection form is presented to the user (step 305), the parent candidate p selected by the user is input (step 306), and the above process is performed. repeat. If the stratification process by the user is not continued (step 304, N), the stratification result is output (step 307).

  FIG. 10 is a diagram showing a flow of a user interface when performing hierarchization in an embodiment of the present invention.

  (1) A tag hierarchization menu is displayed on the client device 2 of the user, and tag candidates in a hierarchy one level lower than the top tag are displayed as recommended tags.

  (2) Based on the tag selected by the user, the index DB 15 is referenced to further display tag candidates in the next lower hierarchy, and a selection input from the user is accepted.

  In (1) and (2), it is also possible to accept user manual input.

  Next, the query expansion processing unit 50 and the search result display control unit 90 at the time of search will be described.

  FIG. 11 is a flowchart of the query expansion processing unit according to the embodiment of the present invention.

  When the search query q and the search option o are given from the client device 2 as inputs (step 401), the query expansion processing unit 50 stores the index DB 15 in the index DB 15 when the search option o is a conceptual search (step 402, Y). A ranking result based on the stored feature vector distance is output (step 403). If o is not a search concept (step 402, N) and the search option o is a similar tag search (step 404, Y), a set of neighbor tags and q acquired from the neighbor index qT stored in the index DB 15 (Step 405), the tags of qT are combined with OR, and a normal search process is performed by the search engine (step 406), and the search result display control unit 90 outputs the search result based on qT (step 406). On the other hand, when o is not a concept search and a similar tag search (step 402, N, step 404, N), a normal search result by q is output (step 408).

  As described above, the query expansion processing unit 50 includes a concept search and a similar tag expansion search in addition to a normal search process as a query process as an internal process.

  Concept search is a search that ranks based on the distance between a tag and a feature vector to be searched. In general, the number of results is increased when there are few normal search results, and the number of results is decreased when there are too many results.

  Further, the similar tag extended search is a search method that uses a neighborhood index and uses a neighborhood tag of a search tag and ranks the union of a neighborhood tag set and a search tag result.

  The following is an example comparing the prior art with the present invention.

  FIG. 12 is an example of an operation menu for bundling tags created in the prior art and the present invention. Fig. (A) is created by the prior art. The list of tags that can be seen when selecting or creating a bundle name is arranged in alphabetical order or frequency order, and how to select the bundle name. It is difficult to grasp intuitively. FIG. 6B is created using the present invention, and tags are grouped and displayed in a list based on relevance whose meaning has been analyzed, and it is possible to intuitively select tags that are appropriate as bundle names. It becomes possible.

  FIG. 13 shows an example of an operation menu for hierarchizing tags created in the prior art and the present invention, and FIG. 14 shows a lower tag in the operation menu for hierarchizing tags created in the prior art and the present invention. This is an example of selecting. FIG. 13 shows the selection operation of the highest tag that is the start of hierarchization, and FIG. 14 shows the selection operation of the lower tag of the tag “programming” designated as one of the highest tags.

  In the prior art shown in FIGS. 13A and 14A, when a tag in a certain hierarchy is selected, it is necessary to select from a list of existing tags or to specify a keyword arbitrarily by the user. In the present invention shown in FIGS. 13 (B) and 14 (B), when the highest tag is selected, a list of tags that can be determined as having a high abstraction level and frequency among all tags used by the user is displayed. When selecting a lower tag of a tag, a tag having a lower abstraction level and a semantically closer distance than the tag is recommended. As a result, the user can easily arrange tags hierarchically. If you want to place tags other than recommended tags or use arbitrary keywords, use the conventional method.

  FIG. 15 shows an example of a normal search result and a result of query expansion.

  The query is 'xquery', but if it is not a tag that is used so frequently, there is a problem that the number of results is reduced in the prior art. On the other hand, in the present invention, it is possible to display a tag having a meaning similar to 'xquery', and to display a result by adding a semantic relation, which contributes to an improvement in the recall rate of the search result. Of course, there is a part that cannot easily be said to improve the recall rate due to increased noise, but by providing a user interface that interactively shows both the normal result and the result of query expansion, information is acquired for the user It can be linked to giving options to do.

  In the above-described embodiment, the social bookmark system is described as an example among the cooperative classification systems, but any cooperative classification system may be used without being limited to the bookmark.

  In addition, the operation | movement of each component of the cooperative classification apparatus shown in FIG. 4 is built as a program, installed in a computer (server) used as the cooperative classification apparatus, executed, or distributed via a network. Is possible.

  Further, the constructed program can be stored in a portable storage medium such as a hard disk, a flexible disk, or a CD-ROM, and can be installed or distributed in a computer.

  The present invention is not limited to the above-described embodiment, and various modifications and applications can be made within the scope of the claims.

  The present invention can be applied to a system in which a plurality of users classify and supply information such as bookmarks, photographs, videos, books, and papers.

It is a principle block diagram of this invention. It is a figure which shows matching with the cooperative classification system of this invention, and SBM. It is a block diagram of the system outline in one embodiment of this invention. It is a block diagram of the cooperative classification apparatus in one embodiment of this invention. It is a block diagram of the correlation process part and indexing process part in one embodiment of this invention. It is a flowchart of the correlation process in one embodiment of this invention. It is a flowchart of the bundling recommendation process in one embodiment of this invention. It is a figure which shows the flow of the user interface at the time of bundling in one embodiment of this invention. It is a flowchart of the hierarchization recommendation process in one embodiment of this invention. It is a figure which shows the flow of the user interface at the time of performing hierarchization in one embodiment of this invention. It is a flowchart of the query expansion process part in one embodiment of this invention. It is an example of the operation menu which bundles the tag produced by the prior art and this invention. It is an example of the operation menu which performs the hierarchy of the tag produced by the prior art and this invention. This is an example in which a lower tag is selected in the operation menu for creating a tag hierarchy created in the prior art and the present invention. It is an example of a normal search result and a query expansion result.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Information sharing server, cooperative classification apparatus 2 Client apparatus 10 User interface means, user interface 15 Index DB (database)
20 Association Processing Unit, Association Processing Unit, Association Engine 21 Element Extraction Unit 22 Pre-Filter Processing Unit 23 PLSI Processing Unit 30 Indexing Processing Unit, Indexing Processing Unit, Indexing Engine 31 Neighborhood Indexing Unit 32 Grouping Hierarchy Unit 40 Bundled / Hierarchical recommendation means, bundled / hierarchical recommendation processing section 50 Search query expansion means, query expansion section, query expansion engine 60 Classification axis database (DB), database server 65 Database 70 Classification target database (DB), database server 75 User Database (DB)
80 communication means, communication section 90 result switching display means, search result display control section

Claims (6)

A collaborative classification information processing device for each user to classify and provide information such as bookmarks, photos, videos, books, papers, etc.
User interface means as a data input / output unit;
A communication means for performing data communication;
A database that stores tags, images, classification axes including audio, classification targets, and user information classified by a plurality of users;
When text data is input from the user's client device, referring to the contents of the database, conceptually associating the text data, and storing in an index database; and
Indexing processing means for indexing and grouping the associated data of the index database and storing in the index database;
A collaborative classification information processing apparatus characterized by comprising: The association processing means includes
Element extraction processing means for extracting co-occurrence data having a characteristic property from the text data with reference to the database;
Filtering means for removing noise using a predetermined minimum frequency condition of elements in the co-occurrence data;
Vector calculation means for obtaining a probability vector of each element of the co-occurrence data from which noise has been removed;
The cooperative classification information processing apparatus according to claim 1, including: The indexing processing means includes
Neighborhood index assigning means for indexing an element close to the element designated by the user as a neighborhood index;
Group index assigning means for indexing an element belonging to the same group as the element designated by the user as a group index;
Hierarchical index assigning means for indexing a lower element group and an upper element of an element designated by the user as a hierarchical index;
The cooperative classification information processing apparatus according to claim 1, including:   Based on the bundle name input from the user, the group index or the hierarchical index of the index database is transmitted to the client device of the user, and an instruction not to continue the hierarchical processing is input from the user The collaborative classification information processing apparatus according to claim 1, further comprising a bundling / hierarchical recommendation unit that repeats up to. When a search query is input from the user, a normal search is performed based on the search query, and the search query is expanded and searched by referring to the index database based on the search query. Expansion means;
Means for switching and displaying a normal search result and a result searched using the extended search query based on an instruction of the user;
The cooperative classification information processing apparatus according to claim 1, further comprising:   A cooperative classification information processing selection program for causing a computer to function as each means constituting the cooperative classification information processing apparatus according to any one of claims 1 to 4.

Images