JP2009217724A - Relevant document estimation device, relevant document estimation method, program and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a relevant document estimation device, enabling a user having no field knowledge or knowledge of inter-document structure to easily find out a relevant document, a relevant document estimation method, a program and a recording medium. <P>SOLUTION: In the relevant document estimation device 100, an importance propagation execution part 130 calculates importance for each of topics and documents in a topic/document graph by performing propagation of importance in the topic/document graph by use of importance propagation probability tables 131-133 showing a propagation probability of importance between adjacent topics or documents in the topic/document graph, and generates a topic/document graph with importance in which the calculation result of importance is assigned to the topics and documents. An importance adjustment part 140 adjusts the numerical values in the calculation result of importance for the topics or documents. A relevant document presentation part 150 presents a link group 220 to relevant documents on a display screen 200 based on the adjusted topic/document graph with importance. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a related document estimation apparatus, a related document estimation method and a program, and a recording medium for estimating a related document to be presented to a user.

  Metadata is information given to content and data. As an example of metadata given to a document, for example, there is information on the content of the document such as a document title and a topic. Ontology is a collection of definitions related to things, descriptions of relationships between things, and vocabulary definitions for describing them.

  Document metadata describes the topic of each document using the vocabulary defined in the ontology, and expresses the relevance of the topic of the document using an ontology graph, or searches for documents with related topics. It becomes possible to do.

  Patent Document 1 describes a related document acquisition method by pattern matching. In the method described in Patent Document 1, after storing a registered document, partial structure data is generated for each registered document, and related portions in the document and between documents are extracted from the partial structure data of all documents. The method described in Patent Document 1 applies a grammatical connection relationship acquisition rule to the extracted related portion, and acquires a semantic relationship that matches the rule. Furthermore, in the method described in Patent Document 1, link information is assigned to a document having an associated part pair or a document pair, and a link list as original data presented to the user is generated for each part to which link information is assigned.

  Patent Document 2 describes a method for modeling document access in a collection based on usage information obtained for a collection of linked documents. The method described in Patent Document 2 generates an attenuation function for attenuation as activity propagates through the network, and executes the activation propagation algorithm to add activity to a set of documents of interest. Its activity propagates through the network until it reaches the activation stability pattern across all documents. Then, target usage information is extracted from the active pattern.

  FIG. 23 is a block diagram showing a configuration of a related document estimation apparatus using a conventional ontology.

  In FIG. 23, the related document estimation device 10 is configured to include an attention topic selection unit 11, a topic / document graph creation unit 12, an important document extraction unit 13, and a related document presentation unit 14.

  The attention topic selection unit 11 selects the attention topic according to an instruction from the user. The instruction from the user is given by the user's page browsing operation (clicking a link, pressing a button, inputting a character in an input field, etc.). In the attention topic selection unit 11, the user selects a hyperlink to a document whose subject is the corresponding topic, or the user selects the corresponding topic from a list of topics indicated to the user. The attention topic selection unit 11 estimates a user's attention topic from an action performed by the user in search of information related to the topic according to the user's selection.

  Here, the topic refers to things and concepts in general. For example, product classifications such as “high-definition video camera HDC-SD ***”, model names, and device functions such as “dubbing to DVD” can also be topics. The relationship between topics refers to the relationship between topics. For example, when the relationship between topics is what the user wants to achieve on the device (task) or the operation of the device is expressed as a topic, a task is a subordinate task of another task, or a certain operation An operation for realizing a certain task. The attention topic is a topic estimated by the attention topic selection unit 11 that the user will be interested in at present. Inferred by the user entering in an input field or selecting a specific topic from a topic list.

  The topic / document graph creation unit 12 creates a topic / document graph from the topic of interest using the document set 15 with metadata, the field ontology 16 and the document ontology 17. The topic / document graph is a graph representing the relationship between topics, the relationship between documents, and the relationship between topics / documents. The topic / document relationship is a relationship between a topic and a document. For example, a topic being the subject of a document is one of topics / document relationships.

  The important document extraction unit 13 extracts an important document from the created topic / document graph using the important document extraction rule 18 and outputs the topic / document graph with importance to the related document presentation unit 14. The importance is a numerical value indicating how important a topic or document is for the user and how much the user needs to show to the user. The importance is obtained as a result of estimation by the system. An important document is a document whose importance exceeds a certain threshold or is within a certain order from the highest importance in a document set. The important document extraction rule 18 is a rule that describes a pattern regarding which document is to be an important document when a topic of interest or a document of interest is given.

  The related document presentation unit 14 presents the extracted topic / document graph with importance. The related document is a document that is currently being browsed or a document that has been determined by the system to be related to a topic that is currently focused on.

  The document set 15 with metadata refers to a set of documents with metadata attached. The metadata may be embedded in the document itself by some markup, or a file describing the metadata may be created separately from the document. The markup language includes Web standard language HTML (Hyper Text Markup Language) and XML (eXtensible Markup Language) that can exchange created content between various information devices such as computers, TVs, and mobile phones. There is.

  The field ontology 16 is an ontology for describing knowledge in a specific field such as information on home appliances. The field knowledge in this case is knowledge of a specific field such as information on home appliances.

  The document ontology 17 is an ontology for describing information regarding a document type, a subject matter of a corresponding document, a partial document, and the like. The document in this case is the same as the “document”. The document is not limited to text, but also includes photographs, moving images, and parts thereof (for example, a paragraph in a text file).

  FIG. 24 is a diagram for explaining related document acquisition by pattern matching. In the figure, the task is that the user intends to realize with the device. The task page is a page describing a task.

As shown in FIG. 24A, a pattern for acquiring a caution page related to a task during browsing of the task page cannot acquire a caution page related to a partial operation of the operation shown in FIG. That is, the attention page cannot be acquired unless the route from the task to the operation, the operation to the partial operation, and the partial operation to the attention is predicted and described in advance.
JP 2006-185334 A JP 11-328226 A J. Pearl: Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference, Morgan Kaufmann, 1988

  However, such related document display devices have the following problems.

  (1) The related document acquisition method based on pattern matching described in Patent Document 1 can only present documents that match a previously described pattern. There is no guarantee that all important relevant documents can be presented.

  For example, in the related document acquisition by the pattern matching shown in FIG. 24, it is difficult to describe a sufficient pattern by predicting various graph structures in advance.

  (2) In the related page guidance that does not use the ontology described in Patent Document 2, only transitions by links between pages are considered, so there are cases where it cannot be reached by links. Also, a page with a large number of hops, which is the number of times to follow a hyperlink, is not recommended.

  (3) The usage guidance on the website has the following problems.

  It is impossible to know where the requested information is, and it is not known in advance whether the information to be requested is included in the link destination. Therefore, even if there is information to be sought, it cannot always be reached. In the first place, the user may not know what he has to know. Domain knowledge and site structure knowledge are required. Site structure knowledge is knowledge about what kind of hyperlinks are made between documents on the Internet website. Users who do not have domain knowledge or knowledge of the structure between documents find a related document. It was difficult.

  Therefore, in the conventional technology, there is a problem that it is difficult for a user who does not have the domain knowledge or the knowledge of the structure between documents to find a related document.

  The present invention has been made in view of the above, and a related document estimation device, a related document estimation method, and a program that can easily find a related document even by a user who does not have domain knowledge or knowledge of the structure between documents. It is another object of the present invention to provide a recording medium.

  A related document estimation apparatus according to the present invention is a related document estimation apparatus that estimates a related document to be presented to a user, and includes an attention topic selection unit that selects a topic that a user is interested in, a document set with metadata, a relationship between topics, and A topic / document graph creation unit that creates a topic / document graph from the topic of interest using inter-document relationships, and an importance propagation probability that represents the propagation probability of the importance of adjacent topics or documents in the topic / document graph Using the table and the importance propagation probability table, the importance of the topic / document graph is propagated to calculate the importance of the topic and the document in the topic / document graph, and the calculated result is used as the topic. Importance propagation execution unit that generates topic / document graph with importance attached to document, and with calculated importance Based on the topic document graph employs a configuration and a related document presenting section for presenting the related document.

  The related document estimation method of the present invention is a related document estimation method for estimating a related document to be presented to a user, the step of selecting a topic to which the user is interested, a document set with metadata, a relationship between topics, and a relationship between documents Using the above, the step of creating a topic / document graph from the topic of interest and the importance propagation probability table representing the propagation probability of the importance of adjacent topics or documents are used to propagate the importance in the topic / document graph. To calculate the importance of the topic and the document in the topic / document graph, and generate a topic / document graph with importance by assigning the calculation result to the topic and the document.

  In addition, the present invention is a program for causing a computer to execute a related document estimation apparatus that estimates a related document to be presented to a user. The computer selects a topic to be noticed by the user, a document set with metadata, a topic Using the inter-relationship and inter-document relation, creating a topic / document graph from the topic of interest, and using the importance propagation probability table representing the propagation probability of the importance of adjacent topics or documents, the topic / document Calculating importance of topics and documents in the topic / document graph by propagating importance in the graph, and generating a topic / document graph with importance by assigning the calculation result to the topics and documents; Is a program for executing the above as a related document estimation device.

  In addition, the present invention is a program for causing a computer to execute a related document estimation apparatus that estimates a related document to be presented to a user. The computer selects a topic to be noticed by the user, a document set with metadata, a topic Using the inter-relationship and inter-document relation, creating a topic / document graph from the topic of interest, and using the importance propagation probability table representing the propagation probability of the importance of adjacent topics or documents, the topic / document Calculating importance of topics and documents in the topic / document graph by propagating importance in the graph, and generating a topic / document graph with importance by assigning the calculation result to the topics and documents; Adjusting the numerical value of the importance estimation result for the calculated topic or document. If, based on the adjusted with importance topic document graph is a program for executing a step of presenting the related document as related document estimator.

  According to the present invention, in the topic / document graph, the probability that each node is important is propagated, and the related knowledge is presented based on the calculated topic / document graph with importance. Even users who do not have knowledge of the structure between documents can easily find related documents.

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

  FIG. 1 is a diagram showing a configuration of a related document estimation apparatus according to an embodiment of the present invention. The related document estimation apparatus according to the present embodiment will be described as an example applied to a system that is plugged into a user's personal computer and that allows a user to access a website and find a related document.

  In FIG. 1, the related document estimation apparatus 100 includes an attention topic selection unit 110, a topic / document graph creation unit 120, a document set database (DB) 121 with metadata, a topic relationship DB 122, a document relationship DB 123, and importance propagation execution. Section 130, topic importance propagation probability table 131, document importance propagation probability table 132, topic / document importance propagation probability table 133, importance adjustment section 140, and related document presentation section 150.

  The attention topic selection unit 110 selects, as an attention topic, a topic that the user is interested in according to an instruction from the user. The instruction from the user is the user's page browsing operation (clicking a link, pressing a button, inputting a character in an input field, etc.). This is also the topic selection by the user. In the selection of a topic by the user, the user selects a corresponding topic from a selection of a hyperlink to a document whose subject is the corresponding topic or a list of topics shown to the user.

  Topics also refer to things and concepts in general. For example, topics include “high-definition video camera HDC-***”, “dubbing to DVD”, and the like. The relationship between topics refers to the relationship between topics. For example, the relationship between topics is that a certain task is a subordinate task of another task, or an operation for realizing a certain operation. An operation for realizing a task with a certain operation is a relationship between topics. When a task 1 is a special case of task 2, task 1 is said to be a task lower than task 2. For example, the task “Record to DVD” is a subordinate task of the task “Record”. The above tasks are what the user wants to do. The topic of interest is a topic that is presumed that the user is currently interested. It is estimated by the user inputting in the input field or selecting from the topic list.

  In this specification, an operation refers to an action that a user performs on a home appliance. From basics such as pressing a button and turning the jog dial to a series of basic operations such as “playing a DVD”, it is widely called “operation”. A partial operation refers to an operation that constitutes a series of operations. “Insert DVD” is a partial operation of “Play DVD”.

  The topic / document graph creation unit 120 creates a topic / document graph from the topic of interest using the document set DB 121 with metadata, the inter-topic relationship DB 122, and the inter-document relationship DB 123. Here, each node has a class (topic, document, etc.) / Identifier (task, etc.). The topic / document graph creation unit 120 links nodes using classes / identifiers, and links and associates each node with the importance propagation probability tables 131 to 133. The topic / document graph is a graph representing the relationship between topics, the relationship between documents, and the relationship between topics / documents. The topic / document relationship is a relationship between a topic and a document. For example, a topic being the subject of a document is one of topics / document relationships. As will be described later, the graph is composed of nodes and branches. A node is one of the basic elements that make up a graph together with branches. In the topic / document graph, each topic or document is a node. A branch is a line connecting nodes in the graph, and represents a direct relationship between the nodes. When the above relationship has a direction, it is assumed that the branch representing the appearance system also has a direction. This is called a directional branch. In the graph, the directional branch is represented by an arrow connecting the nodes. In the description of this embodiment, “topic or document” means “topic and / or document”. This is to avoid complexity in description. Accordingly, “topic or document” includes “topic and document”. Also included are topics or documents.

  In this embodiment, it is assumed that all relationships have a direction. In the graph, a series of branches connecting two nodes is called a path. For example, when there is a branch E1 from the node A to the node B, and there is a branch E2 from the node B to the node C, the row of branches (E1, E2) is a path from A to C.

  The metadata-added document set DB 121 stores a set of documents to which metadata is attached. The metadata may be embedded in the document itself by some markup, or a file describing the metadata may be created separately from the document. A character string attached to a hyperlink is called a link character string. In a typical example of a link character string, a document linked by a corresponding hyperlink is displayed by clicking a portion where the character is displayed.

  The inter-topic relationship DB 122 accumulates relationships between topics. For example, the relationship between topics is that a certain task is a subordinate task of another task, or an operation for realizing a certain operation. When a task 1 is a special case of task 2, task 1 is said to be a task lower than task 2. For example, a task “record on DVD” is a subordinate task of a task “record”.

  The inter-document relationship DB 123 stores the relationship between documents. The inter-document relationship includes a partial document relationship, a relationship between the revised version and the original version, a relationship between the photograph and the alternative text, and the like.

  The importance degree propagation execution unit 130 propagates the importance degree in the topic / document graph using the importance degree propagation probability representing the propagation probability of the importance degree of the adjacent topic or document in the topic / document graph. The importance level propagation execution unit 130 calculates the importance level of the topic and the document in the topic / document graph, and generates the topic / document graph with importance level assigned to the topic and the document based on the calculation result of the importance level. That is, the importance propagation executing unit 130 sets the importance of the topic / document of interest, and performs a propagation calculation between nodes to generate a topic / document graph with importance. The importance level propagation execution unit 130 outputs the generated topic / document graph with importance level to the importance level adjustment unit 140. The importance is a numerical value indicating how important a topic or document is to the user and how much the user needs to show the user. The importance is obtained as a result of estimation by the system. An important document is a document whose importance exceeds a certain threshold or is within a certain order from the highest importance in a document set. Importance of importance means that the importance of a topic or document in the topic / document graph affects the importance of adjacent topics or documents, and the importance of the topic or document changes accordingly. This is one of the features of this embodiment.

  The inter-topic importance propagation probability table 131 stores the inter-topic importance propagation probability, the inter-document importance propagation probability table 132 stores the inter-document importance propagation probability, and the inter-topic importance propagation probability table 133. Stores the topic-document importance propagation probability. It is one of the features of this embodiment that the importance propagation probability tables 131 to 133 are provided. The importance propagation probability tables 131 to 133 indicate the probability that an adjacent topic or document is important (or not important) as a conditional probability for a combination of a topic or a document importance that is a parent in a topic / document graph. It is the table shown. A parent is said to be the parent of B when there is a directed branch that leaves node A and enters node B in the graph. This is the opposite concept of “child”. A child is said to be a child of A when there is a directed branch that leaves node A and enters node B in the graph. The directional branch is described at the branch. Here, in the present embodiment, the importance propagation probability is described as the importance propagation probability table. It only needs to indicate “importance propagation probability indicating the propagation probability of the importance of adjacent topics or documents in the topic / document graph”, and it should be “table” as well as “table”. Absent. That is, any importance propagation probability that can be associated is included in the importance propagation probability table described in the claims of the present application. The synthesis of the importance propagation probability table will be described later with reference to FIG.

  The importance level adjustment unit 140 adjusts the numerical value of the importance level estimation result for the topic or document. The importance adjustment is to adjust the numerical value of the importance estimation result for the topic or document, and is one of the features of this embodiment.

  The related document presentation unit 150 presents a related document based on the calculated topic / document graph with importance. The related document is a document that is currently being browsed or a document that is determined by the system to be related to a topic of interest.

  The related document estimation apparatus 100 is configured by a server computer, for example. Further, it is realized by a plug-in program to a personal computer owned by the user. When the user plugs in the personal computer, the computer can realize the related document presentation. This plug-in may be provided through the Internet. A computer such as a personal computer includes a CPU, a storage device such as a ROM / RAM, an external storage device such as an HDD, an input device, a display device, an input / output interface, and a communication interface.

  In the related document estimation apparatus 100 when plugged in to a personal computer, the attention topic selection unit 110, the topic / document graph creation unit 120, the importance level propagation execution unit 130, and the importance level adjustment unit 140 are processed by a related document presentation process by the CPU. Realized. Further, the inter-topic relationship DB 122, the inter-document relationship DB 123, the importance level propagation execution unit 130, the inter-topic importance level propagation probability table 131, the inter-document level importance probability table 132, and the topic / inter-document importance level propagation probability table 133 are stored. Corresponds to or is stored in the device. The related document presentation unit 150 corresponds to a display device. The CPU controls the operation of other components in the related document estimation device 100 and also controls the data flow between the components of the related document estimation device 100 as necessary. In addition, necessary programs and / or data that implement the functions of the related document estimation apparatus 100 are stored in a storage device, and data in various processing stages are stored in a RAM.

  Hereinafter, the operation of the related document estimation apparatus 100 configured as described above will be described.

  FIG. 2 is a flowchart showing the overall operation of the related document estimation apparatus 100. As described above, it is executed by the CPU as a related document presentation program. In the figure, S is each step of the flow.

  In step S11, the topic / document graph creation unit 120 creates a topic / document graph from the topic of interest using the document set with metadata, the relationship between topics, and the relationship between documents.

  In step S12, the attention topic selection unit 110 sets the attention topic or the attention document and sets the initial value of the importance. The topic of interest is a topic that is presumed that the user is currently interested. The topic of interest is estimated by the user entering in the input field or selecting from the topic list. The document of interest is a document that is estimated to be viewed by the user at present. Further, the initial state is a state in which the topic of interest or the document of interest is not given.

  In step S <b> 13, the importance degree propagation execution unit 130 performs importance degree propagation using the inter-topic importance degree propagation probability table 131, the inter-document importance degree propagation probability table 132, and the topic-to-document importance degree propagation probability table 133. Specifically, the importance propagation execution unit 130 uses the inter-topic importance propagation probability table 131, the inter-document importance propagation probability table 132, and the topic / inter-document importance propagation probability table 133, in the topic / document graph. Calculate importance. The importance level propagation execution unit 130 calculates a topic / document graph with importance level that propagates from the topic of interest to the corresponding document based on the estimation result. A specific calculation method of importance propagation will be described later.

  In step S14, the importance level adjustment unit 140 performs importance level adjustment for adjusting the numerical value of the importance level estimation result for the topic or document.

  In step S <b> 15, the related document presenting unit 150 presents the related document based on the topic / document graph with importance adjusted by the importance adjusting unit 140.

  In step S16, the attention topic selection unit 110 determines whether or not an instruction from the user has been accepted. The instruction from the user is a user's page browsing operation, for example, clicking a link, pressing a button, or inputting a character in an input field. If an instruction from the user is accepted, the process returns to step S12 and the above process is repeated. If an instruction from the user is not accepted, the instruction from the user is waited.

  FIG. 3 is a diagram for explaining an example of topic / document graph creation by the topic / document graph creation unit 120. 3A shows the relationship between topics, FIG. 3B shows the relationship between topics and documents, and FIG. 3C shows the relationship between documents. FIG. 3D shows an example of creating a topic / document graph after merging the above-described topic relationship, topic / document relationship, and document relationship.

  In each figure to be described later including this figure, an ellipse in the figure represents a topic, an operation, a task, and a node, and a rectangle in the figure represents a document and a page. Lines connecting ellipses or rectangles indicate the respective relationships.

  The relationship between topics shown in FIG. 3A indicates the relationship between topics 1 to 2 and topic 3, where topic 1 and topic 3 are relationship TTa, and topic 2 and topic 3 are relationship TTb. This inter-topic relationship is accumulated in the inter-topic relationship DB 122.

  The topic / document relationship shown in FIG. 3B indicates the relationship between the documents 1 and 2 and the topics 1 and 2. The relationship between document 1 and topic 1 is relationship TDa, the relationship between document 1 and topic 2 is relationship TDb, and the relationship between document 2 and topic 2 is relationship TDa. This topic-document relationship includes that a topic becomes the subject of a document. A subject means that when document D describes topic T, T is the subject of D. In this embodiment, since a topic becomes the subject of a document, there is no limit as to what percentage of the document must be a description about the topic. As for the subject, one document may have a plurality of topics as a subject, or one topic may be the subject of a plurality of documents. The topic / document relationship is stored in the document set DB 121 with metadata.

  The inter-document relationship shown in FIG. 3C indicates the relationship between the document 1 and the document 2, and the document 1 and the document 2 are the relationship DDa. This inter-document relationship is accumulated in the inter-document relationship DB 123.

  The topic / document graph creation unit 120 merges the topic relationship, the topic / document relationship, and the document relationship to create a topic / document graph. Topic 1 and topic 2 shown in FIG. 3 (A) are shown in topic 1 and topic 2 shown in FIG. 3 (B), and document 1 and document 2 shown in FIG. 3 (B) are shown in FIG. 3 (C). The topic / document graph shown in FIG. 3D is created by merging with document 1 and document 2 respectively.

  As shown in FIG. 3D, in the created topic / document graph, branches from document 2 to document 1 are connected by a relationship DDa, and branches from document 2 to topic 2 are connected by a relationship TDa. In FIG. 3D, branches from document 1 to topics 1 and 2 are connected by relationships TDa and TDb, respectively. Then, branches from topic 1 to topic 3 are connected by the relationship TTa, and branches from topic 2 to topic 3 are connected by the relationship TTb.

  Next, a specific example of the topic / document graph will be described with reference to FIGS.

  The topic is an AV device such as a video camera or a DVD, for example, “video camera HDC-***”, “dubbing to DVD”, or the like.

  FIG. 4 is a diagram for explaining an example of a topic / document graph. As described above, ellipses in the figure indicate topics, operations, tasks, and nodes, and rectangles in the figure indicate documents and pages.

  The topic / document graph in FIG. 4 is created by the same method as the topic / document graph in FIG. 3D created by merging the relationships in FIGS. However, in FIG. 4, in order to clarify each relationship, the display of the arrow indicating the route is changed. That is, in FIG. 4, (A) the relationship between topics is indicated by a normal arrow, (B) the relationship between topics and documents is indicated by a thick hatching arrow, and (C) the relationship between documents is indicated by a thin hatching arrow. The same applies to FIGS. 5 to 7 described later.

  In FIG. 4, a case where “DVD recorder electronic manual” is a document of interest is taken as an example.

  The “DVD recorder electronic manual” has “a page about DVD compatibility” and “a page about how to hold a DVD” as partial documents. Here, the partial document refers to the document A that is a partial document of the document B when the document A is included in the document B. For example, the partial document is a case where the document B is a Web page and the document A is one paragraph in the document B. In addition, even when the relationship between the documents is abstract such that the document B is a “manual” made up of a plurality of Web pages and the document A is a file representing one of the pages, the partial document I think that.

  First, the “caution page regarding DVD compatibility” and the “caution page regarding how to hold a DVD” have “caution” as the value of the “document type”. The document type indicates the type of document such as “manual”, “catalog”, “attention”, “important notice”, and the like. Up to this point, the inter-document relationship has been described.

  Next, the “caution page regarding DVD compatibility” and the “caution page regarding how to hold a DVD” each have a topic as the value of the relationship “theme”. Here, when the document D describes the topic T, T is said to be the subject of D. There is no limit as to what percentage of a document must be a description about the topic in order for the topic to be the subject of the document. One document may have a plurality of topics as the subject, or one topic may be the subject of a plurality of documents. In the example of FIG. 4, the “DVD compatibility notice page” has two topics, “Record on DVD with high-definition image quality” and “Play DVD”. In addition, the “Caution page about how to hold a DVD” has the topic “insert DVD” as a subject. This is the end of the explanation of the topic / document relationship.

  The following is a description of the relationship between topics. The topic “Recording on DVD with high-definition image quality” has the value of “theme” in the “caution page regarding DVD compatibility”, and is also a subordinate task of the topic “Recording on DVD”. Here, the subordinate task is said to be a subordinate task of task 2 when task 1 is a special case of task 2. In this case, the task “record on DVD with high-definition image quality” is a subordinate task of the task “record on DVD”.

  “Record on DVD with high-definition image quality” has a node “i” as the value of the relation “operation”, and a “insert DVD” task as the value of the relation “partial operation”. Here, the operation is an action performed by the user on the home appliance. The operations are broadly referred to as “operations”, ranging from basic operations such as pressing a button and turning a jog dial to a series of basic operations such as “playing a DVD”. A partial operation is a partial operation that constitutes a series of operations. “Insert DVD” is a partial operation of “play DVD”.

  “Record on DVD” is a subordinate task of “DVD-related tasks”, while “record on DVD with normal image quality” is a subordinate task. “Record on DVD with normal image quality” has a node j as the value of the relation “operation” and a task “insert DVD” as the value of the relation “partial operation”.

  Next, the topic “play DVD”, which is the other subject of the “caution page regarding DVD compatibility”, will be described. “Play DVD” is a value of the “theme” relationship of “DVD compatibility caution page” and is also a subordinate task of the topic “DVD-related tasks”. “Play DVD” has a node k as the value of the relation “operation”, and further has a “insert DVD” task as the value of the relation “partial operation”. “Insert DVD” is also the value of the relationship “theme” in “Notes on DVD compatibility”.

  FIG. 5 is a diagram for explaining an example of the relationship between topics described using the field ontology. The relationship between topics is extracted from the topic / document graph of FIG. The same components as those in FIG. 4 are denoted by corresponding reference numerals.

  FIG. 6 is a diagram for explaining an example of the relationship between documents described using the document ontology. The relationship between documents is extracted from the topic / document graph of FIG. The same components as those in FIG. 4 are denoted by corresponding reference numerals.

  FIG. 7 is a diagram for explaining an example of the relationship between documents described using the document ontology. The topic / document relationship is extracted from the topic / document graph of FIG. The same components as those in FIG. 4 are denoted by corresponding reference numerals. The subject of FIG. 7 is defined by a document ontology as a relationship representing that a document is written about a topic.

  Next, a specific application example of the topic / document graph will be described.

  FIG. 8 is a diagram for explaining an example of a topic / document graph for home appliances such as a DVD. Ellipses in the figure represent topics, operations, tasks, and nodes, and rectangles in the figure represent documents and pages. Also, an importance propagation probability table is attached to these tasks and pages. The importance propagation probability table is a table representing the probability that a topic or a document is important (or not important) as a conditional probability for a combination of the importance of a topic or a document that is a parent in a topic / document graph. Similar importance propagation probability tables are attached to other tasks and pages.

  FIG. 9 is a flowchart showing the setting process of the topic of interest or the document of interest executed by the topic / document graph creation unit 120. The topic of interest is a topic that is presumed that the user is currently interested. The document of interest is a document that is presumed that the user is currently browsing.

  In step S21, it is determined whether or not the previous user operation specifies a topic. The immediately preceding user operation is an action on the immediately preceding system (clicking a link, selecting an item from a list, inputting in an input field, etc.). Here, the meaning is different from the operation on the home appliance.

  When a topic is directly specified by the previous user operation (for example, when a question is input), in step S23, the importance of the designated topic is set to 1, and the importance of the other displayed document is displayed. Set to 1.

  If the previous user operation does not specify a topic in step S21, it is determined in step S22 whether a document is being displayed. If the document is being displayed, in step S24, the importance level of the document being displayed is set to 1, and this flow ends.

  If the document is not being displayed, the importance of the default topic is set to 1 in step S25, and this flow ends.

  Here, when there are a plurality of designated topics, the importance of each designated topic is set to 1. Also, when the topic specification status is obtained by estimation, a value of 0 to 1 is set as the importance of the candidate topic according to the estimation result.

  FIG. 10 is a diagram for explaining a setting example of a document of interest. As shown in FIG. 10, importance is given to tasks and pages. FIG. 10 shows a state where the importance of the “high-vision recording operation page” is set to 100.0%.

  FIG. 11 is a diagram for explaining an example of the importance degree propagation probability table (topic / document relationship), FIG. 11 (A) shows the topic / document relationship, and FIG. 11 (B) shows the importance degree propagation probability. A table is shown.

  As shown in FIG. 11A, the topic and the document are connected by a relationship “themeOf”. The themeOf is TtheOfD when the topic T is the subject of the document D, and in the topic / document graph, a branch with the label theOfOf is stretched from the node representing T toward the node representing D. In the case of FIG. 11A, the topic is a parent and the document is a child. In the graph, when there is a directional branch that goes out from node A and enters node B in the graph, A is said to be the parent of B, and if it is a directional branch, there can be any number of parents. In the importance propagation probability table (topic / document relationship) in FIG. 11B, the probability that the topic is important / topic is not important and the probability that the document is important / document is not important are parents. It is expressed as a conditional probability for the combination of topic and document importance. The importance degree propagation probability table of FIG. 11B is 1.0 when the probabilities of importance / insignificance of a document that is a child are added from a topic that is a parent.

  FIG. 12 is a diagram for explaining an example of the importance degree propagation probability table (topic / topic relation), FIG. 12 (A) shows the topic / topic relation, and FIG. 12 (B) shows the importance degree propagation probability. A table is shown.

  As shown in FIG. 12A, task 1 and task 2 are connected by the relationship “lower task”. When the task 1 is a special case of the task 2, the task 1 is assumed to be a task lower than the task 2. In the case of FIG. 12A, task 1 is a parent and task 2 is a child. In the importance propagation probability table (topic-topic relationship) in FIG. 12B, the probability that task 1 is important / task 1 is not important and the probability that task 2 is important / task 2 is not important are: It is expressed as a conditional probability for the combination of the topic and the importance of the topic. The importance propagation probability table in FIG. 12B is 1.0 when the probabilities of importance / insignificance of task 1 as a child and task 2 as a child are added.

  Next, the importance level propagation will be described with reference to an explanatory diagram shown in FIG.

  The importance level propagation example will be described with respect to an example of importance level propagation when a high-definition recording operation page is set as a notable document in the topic / document graph for a home appliance such as a DVD shown in FIG. This is a specific example of a method for calculating the importance of a high-definition recording operation, a high-definition recording task, and a high-definition recording task page. Further, regarding the calculation of importance propagation, there is a method such as simulation, or a probability propagation calculation method in a Bayesian network. Here, the method of probability propagation calculation in a Bayesian network is followed. As a condition that can be calculated by this method, the topic / document graph needs to be a graph of a type called Poly Tree. If this is not the case, it is necessary to employ another algorithm such as a simulation method. For example, a method described in Non-Patent Document 1 is known.

  First, the following is defined to explain importance propagation.

  (1) The importance vector is (probability that is important / probability that is not important). Probability of importance is importance. (2) A message is information exchanged between nodes in the propagation of importance. (3) The matrix of the propagation probability table associated with the node means that the importance propagation probability table associated with the node is regarded as a matrix. (4) Normalization means adjusting the sum of elements to be 1 when calculating the importance vector. (5) The normalization term is a constant for normalization.

  FIG. 13 is an explanatory diagram when the high-vision recording operation page is a document of interest in the topic / document graph of FIG. FIG. 13 illustrates only relevant portions of the topic / document graph of FIG. A solid line arrow shown in FIG. 13 indicates an importance vector, a shaded arrow indicates a message from a parent to a child, and a hatching arrow indicates a message from a child to a parent. Xn is an ID assigned to a node. Since this ID is attached to the entire topic / document graph of FIG. 8, it is discontinuous with X2, X4, X8, and the like.

[Calculation example of importance of X8 by probability propagation when importance of X11 is set to 1]
For each node, a vector obtained by normalizing the products of the message components from all the parent nodes and child nodes of the corresponding node is set as the importance vector value of the node.

  The importance vector of X8 is obtained by normalizing the product for each component of the message from X4 to X8 and the message from X11 to X8.

  When the node whose importance is set as the topic / document of interest is only X11, when the importance vector of X11 is set to (1.0 0.0), X4 to X8 in the initial state with respect to X4 to X8 A message equal to the message sent to is sent. This is (35.8 64.2) here.

Further, at this time, X11 to X8 are obtained from the row where “X11 is important” in the importance propagation probability table assigned to X11, that is, the following equation (1) (0.9 0. 1) is sent as a message.
(1.0 0.0) * (0.9 0.1) = (0.9 0.1) (1)

  As a result, the importance vector of X8 is normalized (0.834 0.166) by the value obtained by the following equation (2).

(35.8 64.2) * (0.9 0.1) = (32.22 6.42) (2)
Therefore, the importance of X8 is normalized to be 0.834 (= 83.4%).

[Calculation example of importance of X4 by probability propagation when importance of X11 is set to 1]
At this time, a message (0.74 0.14) obtained by multiplying the matrix of the probability table associated with X8 by the message (0.9 0.1) sent from X11 to X8 from the left is assigned to X8 to X4. Sent. That is, the message sent from X11 to X8 is obtained by the following equation (3).

また、このとき、Ｘ２からＸ４へは初期状態におけるＸ２からＸ４へのメッセージと等しいメッセージが送られている。いま、このメッセージを（０．４１１ ０．５８９）とする。

At this time, a message equal to the message from X2 to X4 in the initial state is sent from X2 to X4. Now, let this message be (0.411 0.589).

  Also, (1.0 1.0) is sent as the message from X7 to X4.

  Thereby, the importance vector of X4 is calculated | required by following Formula (4).

(0.74 0.14) * (0.411 0.589) * (1.0 1.0) = (0.304 0.082) (4)
The importance of X4 is obtained as (0.787 0.213) by normalization.

[Calculation example of importance of X7 by probability propagation when importance of X11 is set to 1]
The message from X4 to X7 is obtained by multiplying the message from X2 to X4 (0.411 0.589) and the message from X8 to X4 (0.74 0.14) for each component (0.304 0. 082) is sent as a message from X4 to X7.

Ｘ７は葉節点（子を持たない節点）であるので、Ｘ７の重要度ベクトルは、Ｘ４からＸ７（０．３０４ ０．０８２）にＸ７に付随する伝播確率表の行列に左から掛けて、正規化したものである。すなわち、Ｘ７の重要度ベクトルは、次式（５）のように（０．２８２ ０．１０５）となる。Ｘ７の重要度ベクトルは、正規化することにより（０．７２９ ０．２７１）となり、Ｘ７の重要度は、０．７２９（７２．９％）となる。

Since X7 is a leaf node (a node having no children), the importance vector of X7 is normalized by multiplying the matrix of the propagation probability table associated with X7 from X4 to X7 (0.304 0.082) from the left. It has become. That is, the importance vector of X7 is (0.282 0.105) as shown in the following equation (5). The importance vector of X7 becomes (0.729 0.271) by normalization, and the importance of X7 becomes 0.729 (72.9%).

  FIG. 14 is a diagram for explaining an initial state of the topic / document graph of FIG.

  As described above, the initial state is a state in which an attention topic or an attention document is not given. In other words, the user is not paying attention to the system. Also, the topic of interest is a topic that is presumed that the user is currently interested. It is estimated by the user inputting in the input field or selecting from the topic list. The document of interest is a document that is estimated to be viewed by the user at present.

  The topic / document graph in the initial state shown in FIG. 14 is the same as the topic / document graph of FIG. 8, the topic relationship, the document relationship, the topic / document relationship, and the topic, operation, task, Nodes, branches, paths, documents, pages, etc. are the same.

  15 and 16 are diagrams for explaining the importance level propagation results of the topic / document graph of FIG. FIG. 15 shows the importance level propagation result when “high-vision recording operation page” is the document of interest in the topic / document graph of FIG. FIG. 16 shows the importance level propagation result when “DVD insertion operation page” is the document of interest in the topic / document graph of FIG. Next, it will be described that when the document of interest changes in the topic / document graph, the importance propagation result changes even in the same topic / document graph.

  In FIG. 15, among the documents in the topic / document graph, those other than the document of interest that have an importance of 60.0% or more are enclosed in blocks, and these are referred to as related documents. This related document will be presented to the user.

  The “high-definition recording operation page” is a document of interest that is estimated to be viewed by the user at present. When the “high-vision recording operation page” is the document of interest, the importance level propagation execution unit 130 (FIG. 1) sets the importance level of the “high-vision recording operation page” to 100.0%. For example, the importance level of 38.7% of the “high-definition recording operation page” of the topic / document graph in the initial state of FIGS. 8 and 14 is 100.0% as shown in the “high-vision recording operation page” of FIG. To do. Importance of importance is executed with this as an initial value. The specific method of importance propagation has been described in detail with reference to FIG. For calculation of importance propagation, a method such as simulation may be used, or a probability propagation calculation method in a Bayesian network may be used. Accordingly, the related documents of the importance propagation example of the topic / document graph of FIG. 15 are “DVD related task page”, “DVD recording task page”, “Hi-vision recording task page”, “compatibility caution page”, and “ “Notes on how to hold a DVD”.

  For example, the importance level of the “high-definition recording task page” in the topic / document graph in the initial state of FIGS. 8 and 14 is 42.9%, but the importance level propagation by the importance level propagation execution unit 130 (FIG. 1). The result is 72.9%. Similarly, the importance is calculated for other documents.

  In FIG. 16, among the documents in the topic / document graph, those other than the document of interest with an importance of 60.0% or more are surrounded by blocks and referred to as related documents.

  In the importance level propagation example of the topic / document graph in FIG. 15, the “high-definition recording operation page” is the target document. However, in the importance level propagation example of FIG. 16, the “DVD insertion operation page” is the target document.

  When the “DVD insertion operation page” is the document of interest, the importance propagation execution unit 130 sets the importance of the “DVD insertion operation page” to 100.0%. For example, the importance 63.9% of the “DVD insertion operation page” in the topic / document graph in the initial state of FIGS. 8 and 14 is set to 100.0%. Importance of importance is executed with this as an initial value. As a result, in the case of the importance propagation example of the topic / document graph in FIG. 16, the related documents are “DVD related task page”, “DVD recording task page”, “DVD playback task page”, “compatibility note page”, and This is a “Caution page about how to hold a DVD”. This is because b. In the case of the importance propagation example of the topic / document graph of FIG. Some contents are different from related documents. That is, in the case of the importance level propagation example of the topic / document graph of FIG. 15, in the case of the importance level propagation example of the topic / document graph of FIG. The “reproduction task page” becomes the related document.

  The related document presentation will be specifically described by comparing FIG. 15 and FIG.

  As shown in FIG. 15, when the document of interest is a “high-definition recording operation page”, pages related to “DVD recording” and “high-definition recording” become related documents due to the propagation of importance. In addition, pages related to “note regarding compatibility” and “DVD operation” are also related documents.

  As shown in FIG. 16, when the document of interest is changed to “DVD insertion operation page”, the pages related to “DVD operation” and “DVD playback” become related documents by recalculation of the propagation of importance. . In addition, pages related to “notes regarding compatibility” and “DVD recording” are also related documents. In FIG. 15, the “high-definition recording operation page” that is the noticed document and the “high-definition recording task page” that is the related document are not related documents.

  Here, in the present embodiment, when the importance propagation execution unit 130 calculates the importance of the topic / document shown in FIGS. 15 and 16 as a result of the importance propagation, the calculated importance is: Documents with a predetermined threshold value (here 60.0%) or more are used as related documents. This related document will be presented to the user. It goes without saying that the above threshold value for determining a related document based on the calculated importance is an example and can be arbitrarily set. Further, the threshold used for the determination may be dynamically changed according to the user's history and other usage situations. For example, when a related document is selected more than a predetermined number in a topic / document graph, it is conceivable that the effect of presenting an effective related document to the user is reduced. In this case, by raising the threshold value, a more important related document is presented to the user.

  In the present embodiment, an example of changing the importance of a topic / document graph using the setting of a document of interest as a trigger has been shown. However, the importance of a topic / document graph is not linked to the setting of a document of interest. It may be changed.

  In addition, the importance regarding the past attention document / topic may be left.

  Next, the importance level adjustment process of the importance level adjustment unit 140 will be described.

  First, the concept behind the importance adjustment is described.

  When a user selects a topic, a document having a specific relationship in a graph structure with a topic group as an option includes important information that affects the selection. Also, documents in different relationships do not contain information that affects selection. Also, documents that affect user selection should be presented in preference to documents that are not.

  FIG. 17 is a diagram for explaining importance adjustment by the importance adjustment unit 140. In the figure, a circle indicates a topic, a hatched circle indicates a topic of interest, a block indicates a document, and a hatched block indicates a document in an effective range.

  In the graph, when there is a directional branch E1 having a label R exiting from the node A and entering the node B and a directional branch E2 having a label R exiting from the node A and entering the node C in the graph, The node A becomes a branch node in a route including E1 and E2. The number of branch nodes in the present embodiment may exceed one, but the importance is lowered when it exceeds two or more.

  If two or more branch nodes must be passed from each document to the topic of interest or the document of interest, the importance of the document is multiplied by C (C is a constant greater than or equal to 0 and less than 1).

  In the branch E (X, Y) included in the path P, with respect to the relation R corresponding to E, there is a child node of Y that is a sibling of X, and a node with a document attached to its descendants including itself In addition, Y is a branch node in the path P. Siblings are said to be siblings when there are directional branches from node A to each of nodes B and C in the graph. Also, if a descendant can reach node B by following some directed branches from node A in the graph according to the direction of the branch, node B is said to be a descendant of node A.

  In FIG. 17A, when the relationship between the topic T1 and the topics T2, T3, and T4 is R1, the topic T1 is a branch node, but the documents A and B both exceed only one branch node. Because it is not included in the effective range.

  In FIG. 17B, when the relationship between the topics T1 and T2 and T3 is R1, and the relationship between the topics T3 and T4 and T5 is R2, the topics T1 and T3 are branch nodes. Document A is included in the effective range because it exceeds only one branch node, but document B is not included in the effective range because it exceeds two branch nodes.

  Further, the document related to all may be made less important. For example, if the document A is related to all of the topics T2, T3, and T4 in FIG. 17A, the importance may be increased by 0.3.

  FIG. 18 is a diagram for explaining the adjustment of the importance of the topic / document graph of FIG.

  In FIG. 18, among the documents in the topic / document graph, those other than the document of interest that have an importance of 60.0% or more are enclosed in blocks, and these are referred to as related documents. Further, the branch node is represented by a thick solid line ellipse, and the directed branch of interest is represented by a thick solid line.

  The “DVD recording task page” is a document of interest estimated to be viewed by the user at present. When the “DVD recording task page” is the document of interest, the importance propagation executing unit 130 sets the importance of the “DVD recording task page” to 100.0%. As a result of the propagation of importance, related documents are determined. In the case of the importance propagation example of the topic / document graph of FIG. 18, the related documents are “a caution page regarding compatibility” and “a caution page regarding how to hold a DVD”.

  The “DVD recording task” is a branch node. Although not shown in the figure, assuming that there are other partial operations, the “DVD insertion operation” is also a branch node.

  The importance level adjustment unit 140 adjusts the importance level for the topic / document graph with importance.

  Since the related document “compatibility note page” has only one branch, the importance is left unchanged. Another related document, “Caution on how to hold a DVD”, exceeds two branch nodes and is not within the valid range, so this importance is multiplied by 0.3.

  As described above, the importance level adjustment unit 140 performs adjustment to lower the importance level when two or more branch nodes are exceeded. As a result, the importance of the nodes that do not affect the current selection of the user is adjusted to be low.

  19 and 20 are diagrams showing examples of related document display screens by the related document estimation apparatus 100, which are display examples of a browser connected to the Internet via a communication interface (not shown).

  19 and 20, the related document presentation unit 150 of the related document estimation apparatus 100 has a display device (not shown), and displays various information such as related documents on the display screen 200 of the display device. On the display screen 200, the target document 210, which is an arbitrary web page that the user is browsing on the right half of the screen, the link collection 220 to the related document estimated on the left half of the screen, and the user directly specifies the topic at the top A question input box 230 is displayed.

  The link collection 220 to the estimated related documents is a collection of hyperlinks to the documents estimated as related documents by the system. Since the document of interest 210 in FIG. 19 is a high-definition video camera page, the links 220 to the estimated related documents are related documents related to video camera operations (eg, taking, viewing, dubbing, editing, etc.). It has become.

  20 is a page related to a high-definition video recording operation, the link collection 220 to the estimated related document is used for DVD (for example, recording to DVD, recording to DVD with high-definition, DVD Compatibility, DVD insertion, DVD holding, etc.). In these related documents, the importance calculated as a result of the importance propagation is higher than a predetermined threshold (for example, 60.0%). Therefore, when the user arbitrarily selects a topic of interest, a related document in accordance with the user's intention is presented.

  The link collection 220 to the estimated related document is displayed on the left side of the display screen 200 and is generated by a browser plug-in. The link character string is a character string attached to the hyperlink. Typically, a document linked by a corresponding hyperlink is displayed by clicking on a place where the character is displayed.

  The user can directly specify a topic by inputting a question item in the question input box 230 and clicking the execution button 230a.

  In the link collection 220 to the estimated related document, the important related document is highlighted by changing a color / bold font or the like, and an item to be particularly noted (“DVD compatibility” in FIG. 19 is “ “Note” mark is displayed, in the case of FIG. 19, “SD card compatibility” and “DVD compatibility”, and in the case of FIG. “Notes on how to hold a DVD” is highlighted.

  FIG. 21 is a diagram showing a link character string when a related document is presented.

  As shown in FIG. 21, the value of metadata (title) is used as a link character string on the display screen 200 for the topic / document graph page. For example, the “caution page regarding DVD compatibility” in the topic / document graph of FIG. 18 is a link character string of the title “note regarding DVD compatibility”.

  In this way, when presenting a related document, a character string attached as a metadata (title) value to the document is used as a link character string.

  The related document presentation results described in detail with reference to FIGS. 1 to 18 are reflected in the link collection 220 to the estimated related documents on the display screen 200 of FIGS. 19 and 20.

  FIG. 22 is a diagram for explaining a synthesis example of the importance propagation probability table. 22A shows an importance propagation probability table when Z has only X parents, and FIG. 22B shows an importance propagation probability table when Z has only Y parents. FIG. ) Shows the importance propagation probability tables when the parents of Z are X and Y, respectively.

  The probability is obtained according to the following equation (5).

p (z = non-important | X, Y) = p (z = non-important | X) × p (z = non-important | Y)
p (z = important | X, Y) = 1-p (z = non-important | X, Y) (5)
When the importance propagation probability table when the parent of Z is only Y shown in FIG. 22A and the importance propagation probability table when the parent of Z is only X shown in FIG. The conditional probability that the child is important is obtained by multiplying each parent value by the conditional probability that the child is important. In addition, the conditional probability that the child is important is obtained by subtracting it from 1.

  As described in detail above, the related document estimation device 100 includes the importance degree propagation probability tables 131 to 133 that represent the propagation probability of the importance degree of the adjacent topic or document in the topic / document graph. The importance degree propagation execution unit 130 propagates the importance degree in the topic / document graph using the importance degree propagation probability tables 131 to 133. Accordingly, the importance propagation execution unit 130 calculates the importance of the topic and the document in the topic / document graph, and generates a topic / document graph with the importance obtained by assigning the calculation result of the importance to the topic and the document. The importance level adjustment unit 140 adjusts the numerical value of the importance level calculation result for the topic or document. Then, the related document presentation unit 150 presents a link collection 220 to the related document on the display screen 200 based on the adjusted topic / document graph with importance. Thereby, an effective collection of links is presented to the user. In this case, as in the conventional example, a variety of graph structures are predicted and sufficient patterns are described, or a route from a task to an operation, an operation to a partial operation, and a partial operation to attention is predicted and described in advance. There is no need.

  Further, since it is not a transition due to a link between pages, the user can surely reach a desired related document. Furthermore, it is possible to find a related document even when it is not known where the requested information is, or when it is not known in advance whether or not the requested information is included in the link destination.

  In this way, the present invention presents related documents based on the topic / document graph with importance, in which the probabilities that each node is important are propagated and calculated. There is an excellent effect that even a user who does not have knowledge of the structure can easily find a related document.

  The above description is an illustration of a preferred embodiment of the present invention, and the scope of the present invention is not limited to this. For example, in the present embodiment, presentation of a related document for home appliances has been described. However, any device and method for estimating a related document to be presented to a user may be used, and the type of related document is not limited.

  In the present embodiment, the names of the related document estimation device and the related document estimation method are used. However, this is for convenience of explanation, and the related document estimation device includes a related document display device, an information presentation device, and a related document estimation method. Of course, the related document display method may be used.

  Furthermore, each part which comprises the said related document estimation apparatus and method, for example, a related document presentation part, the kind of DB, the number, the connection method, etc. may be what.

  The related document estimation device and the related document estimation method described above are also realized by a program for causing the related document estimation method to function. This program is stored in a computer-readable recording medium.

  The related document estimation apparatus according to the present invention can easily find related documents even by users who do not have knowledge of the field or the structure between documents. It can also be applied to Web pages for administrative services.

The figure which shows the structure of the related document estimation apparatus of embodiment of this invention. The flowchart which shows the whole operation | movement of the related document estimation apparatus which concerns on the said embodiment. The figure explaining the example of creation of the topic / document graph by the topic / document graph creation unit of the related document estimation apparatus according to the above embodiment The figure explaining an example of the topic and document graph of the related document estimation apparatus which concerns on the said embodiment The figure explaining an example of the relationship between the topics described using the field ontology of the related document estimation apparatus which concerns on the said embodiment. The figure explaining an example of the relationship between the documents described using the document ontology of the related document estimation apparatus which concerns on the said embodiment. The figure explaining an example of the relationship between the documents described using the document ontology of the related document estimation apparatus which concerns on the said embodiment. The figure explaining an example of the topic * document graph for household appliances, such as DVD of the related document estimation apparatus which concerns on the said embodiment The flowchart which shows the setting process of the attention topic or attention document which the topic and document graph preparation part of the related document estimation apparatus which concerns on the said embodiment performs The figure explaining the example of a setting of the attention document of the related document estimation apparatus which concerns on the said embodiment. The figure explaining an example of the importance propagation probability table | surface (topic / document relationship) of the related document estimation apparatus which concerns on the said embodiment. The figure explaining an example of the importance propagation probability table | surface (topic-topic relationship) of the related document estimation apparatus which concerns on the said embodiment. Explanatory drawing when the HDTV recording operation page is the document of interest in the topic / document graph of the related document estimation apparatus according to the above embodiment The figure explaining the initial state of the topic / document graph of the related document estimation apparatus according to the above embodiment The figure explaining the importance propagation example of the topic and document graph of the related document estimation apparatus according to the above embodiment The figure explaining the importance propagation example of the topic and document graph of the related document estimation apparatus according to the above embodiment The figure explaining importance adjustment by the importance adjustment part of the related document estimation apparatus which concerns on the said embodiment The figure explaining adjustment of the importance of the topic and document graph of the related document estimation apparatus according to the above embodiment The figure which shows the example of the related document display screen by the related document estimation apparatus of the related document estimation apparatus which concerns on the said embodiment. The figure which shows the example of the related document display screen by the related document estimation apparatus of the related document estimation apparatus which concerns on the said embodiment. The figure which shows the link character string at the time of presentation of the related document of the related document estimation apparatus which concerns on the said embodiment. The figure explaining the example of a synthesis | combination of the importance propagation probability table of the related document estimation apparatus which concerns on the said embodiment. The block diagram which shows the structure of the related document estimation apparatus which uses the conventional ontology Diagram explaining related document acquisition by conventional pattern matching

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Related document estimation apparatus 110 Attention topic selection part 120 Topic and document graph creation part 121 Document collection DB with metadata
122 Topic relation DB
123 Inter-document relationship DB
130 Importance Propagation Execution Unit 131 Inter-topic Importance Propagation Probability Table 132 Inter-Document Importance Propagation Probability Table 133 Topic / Document Importance Propagation Probability Table 140 Importance Adjustment Unit 150 Related Document Presentation Unit 200 Display Screen 210 Attention Document 220 Estimation Links to related related documents 230 Question input box

Claims (10)

A related document estimation device for estimating a related document to be presented to a user,
A topic selection section for selecting a topic that the user is interested in;
A topic / document graph creation unit that creates a topic / document graph from the topic of interest using a document set with metadata, a relationship between topics, and a relationship between documents;
In the topic / document graph, the importance propagation probability table representing the propagation probability of the importance of adjacent topics or documents and the importance propagation probability table are used to propagate the importance in the topic / document graph. To calculate the importance of the topic and document in the topic / document graph, and generate an importance-weighted topic / document graph in which the calculation result is assigned to the topic and document,
A related document presentation unit for presenting a related document based on the calculated topic / document graph with importance,
Related document estimation device comprising:   2. The related document estimation apparatus according to claim 1, wherein the importance propagation probability table is a topic importance propagation probability table, a document importance propagation probability table, and a topic / document importance propagation probability table.   The related document estimation apparatus according to claim 1, further comprising an importance adjustment unit that adjusts a numerical value of an importance estimation result for the topic or document calculated by the importance propagation execution unit.   The related document estimation device according to claim 1, wherein the related document presentation unit presents a link collection of related documents based on the generated topic / document graph with importance. A related document estimation method for estimating a related document to be presented to a user,
Selecting a topic that the user is interested in;
Creating a topic / document graph from the topic of interest using a document set with metadata, a topic relationship and a document relationship;
By using the importance propagation probability table representing the propagation probability of the importance of the adjacent topic or document, the importance of the topic and document in the topic / document graph is determined by propagating the importance in the topic / document graph. And generating a topic / document graph with importance by assigning the calculation result to the topic and the document.   6. The related document estimation method according to claim 5, further comprising a step of presenting a related document based on the generated topic / document graph with importance.   The related document estimation method according to claim 5, further comprising a step of adjusting a numerical value of an importance estimation result for the calculated topic or document. A program for causing a related document estimation device to estimate a related document to be presented to a user,
On the computer,
Selecting a topic that the user is interested in;
Creating a topic / document graph from the topic of interest using a document set with metadata, a topic relationship and a document relationship;
By using the importance propagation probability table representing the propagation probability of the importance of the adjacent topic or document, the importance of the topic and the document in the topic / document graph is determined by propagating the importance in the topic / document graph. And a step of generating a topic / document graph with importance, in which the calculation result is assigned to the topic and the document. A program for causing a related document estimation device to estimate a related document to be presented to a user,
On the computer,
Selecting a topic that the user is interested in;
Creating a topic / document graph from the topic of interest using a document set with metadata, a topic relationship and a document relationship;
By using the importance propagation probability table representing the propagation probability of the importance of the adjacent topic or document, the importance of the topic and the document in the topic / document graph is determined by propagating the importance in the topic / document graph. Generating a topic / document graph with importance, in which the calculation result is assigned to the topic and the document;
The step of adjusting the numerical value of the importance estimation result for the calculated topic or document and the step of presenting the related document based on the adjusted topic / document graph with importance are executed as a related document estimation device. Program to let you. A computer-readable recording medium storing the program according to claim 8 or 9.

Images