WO2015056699A1

WO2015056699A1 - Information extraction device and information extraction program

Info

Publication number: WO2015056699A1
Application number: PCT/JP2014/077403
Authority: WO
Inventors: 太輔小田嶋; 健一新城; ヴェトハグェン
Original assignee: 株式会社エーエヌラボ; 太輔小田嶋; 健一新城
Priority date: 2013-10-18
Filing date: 2014-10-15
Publication date: 2015-04-23
Also published as: JP2015079401A; JP5522813B1

Abstract

[Problem] To provide an information extraction device and an information extraction program capable of not only outputting a result, but also presenting information about a reason which has led to the outputting of the result, in order to improve the understanding of how the result was outputted. [Solution] The information extraction device is provided with: a start constituent element specification means for specifying a start constituent element to be extracted; a mapping means for mapping from the start constituent element specified by the start constituent element specification means to an extraction result; a reason extraction means for extracting a reason which has led to the mapping of the extraction result, by the mapping means, from the start constituent element; and an output means for outputting the extraction result mapped by the mapping means and the reason extracted by the extraction means.

Description

Information extraction apparatus and information extraction program

The present invention relates to an information extraction apparatus and an information extraction program for performing an information extraction process. Specifically, the present invention relates to an information extraction apparatus and an information extraction program that output not only the extraction result but also the reason why the result is derived.

A conventional search system simply outputs information obtained by keyword search and filtering as a list, and then leaves it to the user to select. In other words, there are many similar search results, and many similar words are included in the search results. It was a burden on the user side.
In this way, even though the search results are presented, the reason why the search results have come out and the story are not presented, so you can select the one that suits your purpose from the filtered search results It was something that had to be done.

For example, in a data search system for searching for data by keyword, a keyword input means for inputting a keyword, a data input means for inputting data, a data storage means for storing a data feature amount and a keyword together with registered data, and the keyword input A keyword search unit that searches for data by combining keywords input from the unit, a feature amount / keyword extraction unit that automatically extracts keywords and feature amounts from registered data, and a result of search performed by the keyword search unit, There has been provided a data search system including search result narrowing means for narrowing down without using a keyword (see Patent Document 1).

JP 2000-48041 A

In a conventional search system, even if a result or a related keyword is output, the reason until the result is output is not presented. Therefore, there is a problem that it is difficult to judge whether the output result is really meaningful to the user.
Therefore, the problem to be solved by the present invention is not only the output of the result, but also the reason why the result is output for improving the user's satisfaction with the output of the result. It is an object to provide an information extraction apparatus and an information extraction program capable of presenting whether or not it has been reached.

In order to solve the above-mentioned problem, the invention according to claim 1 of the present invention is an information extraction device, wherein a start component specifying means for specifying a start component to be extracted, and the start component specifying means Mapping means for mapping from the start component specified by the extraction result to the extraction result, reason extraction means for extracting the reason why the extraction result is mapped from the start component by the mapping means, and mapping by the mapping means And an output means for outputting the extraction result obtained and the reason extracted by the reason extraction means.

According to a second aspect of the present invention, the mapping means extracts an associative component extraction means for extracting an associative component mapped to the start component specified by the start component specifying means, and the association Goal component extraction means for extracting a goal component mapped to the associative component extracted by the component extraction means, and an extraction result mapped to the goal component extracted by the goal component extraction means It is characterized by comprising a result extracting means for extracting.

According to a third aspect of the present invention, the associative component extracting means is an information table in which an associative component associated with a start component is mapped, and the goal component extracting means is associated with an associative component. The target component is extracted based on the mapped information table, and the output is performed on the basis of at least the associative component among a series of mapped start component, associative component, and goal component associative flow. It outputs by a means.

According to a fourth aspect of the present invention, the information table is prepared for each viewpoint, and the associative component extraction means and the goal component extraction means use only information tables belonging to the same viewpoint. The reason why the data is output by the output means includes the above-mentioned viewpoint.

According to a fifth aspect of the present invention, in the information table used for the associative component extracting means, the associative component associated with the start component is mapped in addition to the associative component associated with the start component. In the same viewpoint, at least one or more other associative components mapped are prepared, and the associative component extracting means is mapped to the start component specified by the start component specifying means. After extracting the associated associative component, using the information table in which other associative components associated with the associative component are mapped, the mapped associative component is extracted, and the goal component extracting means Is mapped to the associative component finally extracted by the associative component extracting means. Characterized in that it is intended to extract the Le component.

According to a sixth aspect of the present invention, a weighting setting unit for which a predetermined weighting is set for each viewpoint, and the priority of each reason is totaled after reflecting the weighting set by the weighting setting unit. A priority totalization unit; and an output rank determination unit that determines an output rank of reasons in descending order of priorities tabulated by the priority tabulation unit, and the output unit outputs a reason for each extraction result, The reason for outputting for each extraction result is output according to the output order determined by the output order determining means.

According to a seventh aspect of the present invention, a computer includes a start component specifying step for specifying a start component to be extracted, and a start component specified by the start component specifying step to an extraction result. A mapping step for performing mapping, a reason extraction step for extracting the reason why the extraction result has been mapped from the start component by the mapping step, an extraction result mapped by the mapping step, and the reason extraction step. An output step for outputting the extracted reason is realized.

According to an eighth aspect of the present invention, in the mapping step, the associative component extracting step of extracting an associative component mapped to the start component specified by the start component specifying step, and the association A goal component extraction step for extracting a goal component mapped to the associative component extracted by the component extraction step, and an extraction result mapped to the goal component extracted by the goal component extraction step. And a result extracting step for extracting.

According to a ninth aspect of the present invention, the associative component extracting step is an information table in which an associative component associated with a start component is mapped, and the goal component extracting step is associated with an associative component. The target component is extracted based on the mapped information table, and the output is performed on the basis of at least the associative component among a series of mapped start component, associative component, and goal component associative flow. It outputs by a step.

In the invention according to claim 10 of the present invention, the information table is prepared for each viewpoint, and the extraction in the associative component extraction step and the goal component extraction step uses only information tables belonging to the same viewpoint. The reason why the data is output by the output step includes the above viewpoint.

According to an eleventh aspect of the present invention, in the information table used in the associative component extracting step, the associative component associated with the start component is mapped in addition to the association component associated with the start component. In the same viewpoint, at least one or more other associative components are mapped, and the associative component extraction step is mapped to the start component specified by the start component specifying step. Extracting the associated associative component, using the information table in which other associative components associated with the associative component are mapped, extracting the associated associative component, and the goal component extracting step Is the association component finally extracted by the association component extraction step. Characterized in that extracts a goal component that is mappings.

According to a twelfth aspect of the present invention, a weight setting step in which a predetermined weight is set for each of the viewpoints, and the priority of each reason is totaled after reflecting the weight set by the weight setting step. A priority totalization step, and an output rank determination step for determining the output rank of the reasons in descending order of priority calculated by the priority totalization step, wherein the output step outputs a reason for each extraction result, The reason for outputting for each extraction result is output according to the output order determined by the output order determining step.

According to the present invention, in order to output not only a simple extraction result but also a series of associative flows from which the extraction result is derived as a reason, the user's satisfaction with the output of the extraction result is improved. There is an effect that can be made.

1 is a block diagram illustrating an internal configuration of an information extraction apparatus 101 in Embodiment 1. FIG. It is the flowchart figure which showed the process sequence which outputs the extraction result in the information extraction device 101 in Example 1, and the reason that the extraction result came to be derived. It is explanatory drawing which showed mapping of the component etc. to the reason which came to derive the extraction result from the start component in Example 1. FIG. FIG. 6 is an explanatory diagram showing an overall system configuration in Embodiment 2. It is the block diagram which showed the internal structure of the information extraction apparatus 101 in Example 2. FIG. It is the flowchart figure which showed the process sequence which outputs the extraction result in the information extraction device 101 in Example 2, and the reason that the extraction result came to be derived. The output example of the reason which came to derive the extraction result in Example 2 and the extraction result is shown.

Next, embodiments of the information extraction apparatus and the information extraction program according to the present invention will be described in detail below with reference to the drawings.

Although not shown, the information extraction apparatus 101 is a computer including a control unit such as a CPU and a storage unit such as a RAM and a ROM.
FIG. 1 is a block diagram showing the internal configuration of the information extraction apparatus 101 in this embodiment.
The specific internal configuration is associated with the target information specifying unit 201 that specifies information to be extracted, the start component specifying unit 202 that specifies the start component from the information to be extracted, the start component of the information table, and the like. An associative component extraction unit 203 that extracts an associative component mapped to the start component based on the mapping information of the component, and an associative component based on the mapping information of the associative component and the goal component of the information table The goal component extraction unit 204 that extracts the mapped goal component, the extraction result mapped to the goal component based on the mapping information of the goal component and the extraction result, and the reason why the extraction result has been derived Extraction result and reason extraction unit 205, extraction results and extraction results are derived An output table 206 that outputs the reason for reaching the screen by screen display, printing, etc., an information table prepared for each viewpoint by mapping the start component and the associative component, and the associative component and the goal component, The information table management unit 207 manages information on the mapping between the goal component and the extraction result.

Next, functions of the information extraction apparatus 101 according to the present invention will be described in detail.
FIG. 2 is a flowchart showing a processing procedure for outputting the extraction result and the reason why the extraction result has been derived in the information extraction apparatus 101 according to the present embodiment.
After the start of implementation (step S101), the information to be extracted is specified by the target information specifying unit 201 (step S102).
The start component specifying unit 202 specifies the start component from the information to be extracted (step S103).
Next, in the information table managed by the information table management unit 207, an associative component associated with the start component is mapped and an information table prepared for each viewpoint is searched. It is determined whether or not the identified start component is hit (step S104).
If the start component is not hit, the determination result is “No”, the process proceeds to step S109, and the process ends.
On the other hand, when the start component is hit, the associative component associated with the start component is mapped and the associative component mapped to the start component is displayed in the information table prepared for each viewpoint. Extracted by the associative component extraction unit 203 (step S105).
Next, in the information table managed by the information table management unit 207, the goal component associated with the association component is mapped and mapped to the association component in the information table prepared for each viewpoint. The goal component extracted by the goal component extraction unit 204 is extracted (step S106).
The result and reason extraction unit 205 extracts and maps the extraction result mapped to the goal component based on the mapping information of the goal component managed by the information table management unit 207 and the extraction result. A series of association flows and viewpoints of the start component, the association component, and the goal component are extracted as reasons (step S107). The start component, the associative component, the goal component, and the viewpoint are not limited to all reasons, and some of them may be extracted as the reason. For example, only the associative component, only the associative component and the viewpoint may be extracted as the reason.
Then, the extraction result and the reason why the extraction result has been derived are output by the output unit 206 (step S108), and the process ends (step S109).

As described above, in this embodiment, after the information to be extracted is specified by the target information specifying unit 201, the start component is specified by the start component specifying unit 202, and from the specified start component to the extraction result. Mapping is performed, the extraction result and the reason why the extraction result is derived are extracted by the result and reason extraction unit 205, and output by the output unit 206.
The mapping from the start component to the extraction result is performed by the associative component extraction unit 203, the goal component extraction unit 204, and the result and reason extraction unit 205.

FIG. 3 is an explanatory diagram showing mapping of components and the like up to the reason why the extraction result is derived from the start component.
In FIG. 3, there are four start components A1, A3, D1, and D3. There are two viewpoints, i and b. There are two information tables AB and BC in viewpoint A, and three information tables DE, EF, and FG in viewpoint B. In the information table AB, B1 associated with A1, B2 and B3 associated with A2, B3 associated with A3, and B4 associated with A4 are mapped.
In the example of the start component A1, since A1 is included in the information table AB, A1 in the information table AB that is associated with the information table AB is first mapped to B1 associated with A1, B1 is mapped to C1 associated with B1 in the information table BC. Here, B1 is an associative component, and C1 is a goal component. Thus, the start component A1, the associative component B1, and the goal component C1 are mapped to form a series of associative flows. Since C1 is mapped to the extraction result X, X is extracted as the extraction result, and the reason why the extraction result X is derived is that the viewpoint 1 and the associative component B1 are extracted as C1 related items. Is done. However, the reason for the extraction result X includes C3 and G1 related reasons as shown in FIG.

<Identification of extraction target information and start components>
First, contents to be extracted information such as newspaper articles, natural language sentences, photos, answers to questions, etc. are prepared. The content of this content is not limited, but it must be able to be dropped to the tag that will eventually become the starting component.
Various methods can be used for specifying the extraction target information by the target information specifying unit 201. For example, if you ask a user and answer "I want to watch a movie", you can select a movie you want to watch from a sample that you have prepared in advance, or you can enter the title of your favorite movie. is there.
Next, the start component specifying unit 202 specifies the start component from the information to be extracted. For example, if the information to be extracted is a newspaper article, a tag may be extracted by text analysis of the newspaper article and specified as a start component.
Furthermore, some start components included in the information table managed by the information table etc. management unit 207 and mapped to the associative components associated with the start components are presented and directly started. The component may be specified by the user.
Here, the identification of the start component means that the meta information attached to the extraction target information is decomposed into text data elements. The start component is configured by a tag that defines meta information of content that is extraction target information.

<Mapping of start components and associative components>
First, an information table (information table AB / information table DE) that is prepared for each viewpoint, such as a place, a landscape, an attraction, a restaurant, etc., mapped with an associative component associated with a start component is searched. Thus, it is determined whether there are start components A1, A3, D1, and D3.
When the start component is hit, an associative component mapped to the start component is extracted. For example, in the case of the start component A1, since the start component A1 exists in the information table AB, B1 mapped to A1 in the information table AB is extracted as an associative component.
In this manner, associative component extraction is performed on the basis of information in the information table that is prepared for each viewpoint and that maps the associative component associated with the start component managed by the information table management unit 207. The unit 203 extracts an associative component mapped to the start component specified by the start component specifying unit 202.

In the information table in which the associative component associated with the start component is mapped, the start component and the associative component are not only mapped 1: 1 but also mapped 1: N. May be. For example, in the information table AB of FIG. 3, B2 and B3 are mapped to A2. If A2 is a shop with a good atmosphere, out of the information tables prepared for each viewpoint, if viewpoint A is the atmosphere, in information table AB, for example, a shop with a beautiful night view of B2, a quiet meal of B3 It may be mapped to two of the shops that can.
However, it is not mapped beyond the viewpoint, and even if it is mapped to a plurality of associative components, the information table in which the start component and the associative component are mapped is not included in the same viewpoint. Since there is only one, it is mapped in the same information table. For example, any of E1 to E4 in the information table DE included in the viewpoint B having a different viewpoint is not mapped to A2 in the information table AB included in the viewpoint i.

In addition, the information table existing from a certain viewpoint includes an information table in which an associative component associated with the start component is mapped, and an information table in which the goal component associated with the associative component is mapped. It is not limited to two. As in the case of viewpoint B, the association between the information table DE to which the associative component associated with the start component is mapped and the information table FG to which the goal component associated with the associative component is mapped. There may be an information table EF in which other associative components associated with the components are mapped. In the case of D1, it is mapped to E1 in the information table DE, but then E1 is mapped to F1 in the intermediate information table EF, and F1 is further mapped to G1 in the information table FG. For example, D1 wants to date, E1 has a good atmosphere, F1 has a beautiful night view, and G1 is a restaurant on a higher floor. In this case, the associative component constituting the reason why G1 is selected is not one associative component, but a set of a plurality of associative components E1 and F1.
In the information table in which other associative components associated with the associative components are mapped, the associative components and the other associative components are not only mapped 1: 1 but also 1 : May be mapped to N. F2 and F3 are mapped to E2 in the information table EF.
As described above, the information table used for extraction of the associative component managed by the information table management unit 207 includes the associative component in addition to the mapping of the associative component associated with the start component. From the same viewpoint, at least one or more may be prepared in which other associative components associated with the above are mapped. In this case, after the associative component extraction unit 203 extracts the associative component mapped to the start component specified by the start component specifying unit 202, another associative component associated with the associative component is displayed. The mapped associative component is further extracted using the mapped information table, and the goal component extracting unit 204 is mapped to the associative component finally extracted by the associative component extracting unit 203. Extract elements.

<Mapping of associative components and goal components>
Based on the information in the information table prepared by each viewpoint and mapping the goal component associated with the association component managed by the information table management unit 207, the goal component extraction unit 204 includes: The goal component mapped to the associative component extracted by the associative component extracting unit 203 is extracted.
In the information table in which the goal component associated with the associative component is mapped, the associative component and the goal associative component are not only mapped 1: 1 but also 1: N. It may be mapped. C1 and C2 are mapped to B2 in the information table BC. For example, this is the case where ribs and loin are mapped to yakiniku.
However, it is not mapped beyond the viewpoint, and even if it is mapped to a plurality of goal components, the information table in which the associative component and the goal associative component are mapped is not included in the same viewpoint. Since there is only one, it is mapped in the same information table.

<Mapping of goal components and extraction results>
The result and reason extraction unit 205 extracts the extraction result mapped to the goal component extracted by the goal component extraction unit 204. More specifically, when the meta information attached to the content decomposed into the goal component and the text data element matches, the content is output as an extraction result. However, content that is obtained by storing mapping information between content as an extraction result and meta information attached to the content in the information table management unit 207 in advance and searching for the mapping information using the goal component as a keyword. May be output as an extraction result.
The extraction result is mapped to the goal component and is not directly mapped to the start component or the associative component.
Further, the goal association component and the extraction result are not limited to the case of 1: 1 mapping. In the case of FIG. 3, the goal component C3 is mapped not only to the extraction result X but also to Y, and the extraction result X is mapped not only to the goal component C3 but also to C1 and G1.

<Output>
The mapping from the start component specified by the start component specifying unit 202 to the extraction result is performed, and not only the extraction result but also the reason that the extraction result is mapped from the start component is the result and reason extraction unit 205. And output by the output unit 206. For example, in the case of extracting a recommended ramen shop based on the information of a user's favorite ramen shop, X is extracted as a result of extracting the ramen shop X as an extraction result. The reason is that the reason is that the store owner of X trained in the same store as the store owner of the ramen shop that the user likes in the past.
In the present invention, the reason for searching for the extraction result, in other words, the connection between the information tables, is more important than searching for the extraction result from among many. Then, when there are three or more information tables in the same viewpoint as in the viewpoint B of FIG. 3, not only the associative component linking the information table DE and the information table EF but also the information table EF and the information table FG are linked. The associative component is also output as the reason.
The output content only needs to include the extraction result and the reason why the extraction result has been derived, and various outputs are possible. In the case of FIG. 3, three extraction results of X, Y, and Z are output. Then, as the reason for X, the viewpoint i and the associative component B1, the viewpoint i and the associative component B3, and the viewpoint b and the associative components E1 and F1 are output. The element B3 and two of viewpoint B and associative components E1 and F1 are output, and as the reason of Z, one of viewpoint B and associative components E3 and F3 is output.
Thus, the reason for output includes a viewpoint, an associative component, and the like. Further, the present invention is not limited to the case where viewpoints, associative components, etc. are simply displayed side by side, and a sentence including these may be generated and output as a reason.
The output result to be output is not limited to text, but may be contents such as newspaper articles, photos, videos, music, and the like. In addition to the content that is the extraction result, content related to the content may be output together.

<Information table>
An information table prepared for each viewpoint by mapping the start component and the associative component, and the associative component and the goal component will be described. The information table in the present invention is a table in which a certain word and an associative word associated with the word are mapped and registered.
An information table exists for each of one or more viewpoints. In addition, the information table management unit 207 manages at least two information tables each mapping the associative component associated with the start component and the goal component associated with the associative component for each viewpoint. .
However, the information table for each viewpoint is not limited to these two, and the information table in which other associative components associated with the associative components are mapped between the two information tables is 1 or Multiple may be included. For example, as in the example shown in FIG. 3, there are two information tables for viewpoint A, but there are three information tables for viewpoint B.

3 is an information table DE in which an associative component associated with the start component is mapped, and an information table in which other associative components associated with the associative component are mapped. There are three EFs, an information table FG to which the goal components associated with the associative components are mapped, but the goal components are from the information table FG to which the goal components associated with the associative components are mapped. Only extracted. That is, one goal component is extracted from the information table FG, but another goal component is not extracted from the intermediate information table EF. Depending on the viewpoint, as shown in FIG. 3B, in addition to the two information tables in which the associative component associated with the start component and the goal component associated with the associative component are respectively mapped, the associative configuration Although there may be an information table mapped between other associative components associated with the element, the goal component is a tag that defines the meta information of the extraction result, so the extraction result is the goal component It is only associated with an element and not directly with an intermediate associative component.
Further, the information table managed by the information table management unit 207 is not limited to the case where the information table is stored and saved in the information extraction apparatus 101. The mapping information itself of the information table is the information of the information extraction apparatus 101. It may exist outside.

In this embodiment, the extraction result and the reason are output reflecting the user's intentionality, the priority of the reason is calculated based on the set weight, and the output order is determined. The point that communication with the terminal 103 is added is different from the case of the first embodiment. Since other parts are the same as those in the first embodiment, description thereof will be omitted as appropriate.
FIG. 4 is an explanatory diagram showing the overall system configuration in the present embodiment.
As shown in FIG. 4, in the information extraction device of the present application, a user terminal 103 installed on the user side is connected to the information extraction device 101 via a communication network 102. In the system configuration shown in FIG. 4, the information extraction apparatus 101 is shown as one and the user terminal 103 is shown as two, but each of these may have a larger number.

The communication network 102 can use various networks such as the Internet, a LAN (local area network), and a WAN (wide area network).

Although not shown, the information extraction apparatus 101 is a computer including a control unit such as a CPU and a storage unit such as a RAM and a ROM.
Although not shown, the user terminal 103 includes a control unit such as a CPU, a storage unit such as a RAM and a ROM, a display unit such as a liquid crystal screen, an input unit such as a keyboard, a communication unit that controls communication with the Internet, and the like. This includes information processing devices including personal computers, smartphones, mobile phones, tablet terminals, PHS, PDAs, dedicated terminals, and the like.

FIG. 5 is a block diagram showing the internal configuration of the information extraction apparatus 101 in this embodiment.
The specific internal configuration is associated with a target information specifying unit 301 that specifies information to be extracted, a start component specifying unit 302 that specifies a start component from the information to be extracted, a start component of an information table, and the like. An associative component extracting unit 303 that extracts an associative component mapped to the start component based on the mapping information of the component, and an associative component based on the mapping information of the associative component and the goal component of the information table The goal component extraction unit 304 that extracts the mapped goal component, the extraction result mapped to the goal component based on the mapping information of the goal component and the extraction result, and the reason why the extraction result has been derived Results and reason extraction unit 305, between viewpoints, between start components, associations Priority aggregation unit 306 that aggregates the priority of extraction results and reasons based on weighting between components or between extraction results, and output rank determination unit 307 that determines output ranks in descending order of the aggregated priorities The output unit 308 for outputting the extraction result and the reason that the extraction result has been derived according to the output order that has been output by screen display or printing, the start component and the associative component, and the associative component and the goal component, respectively Information table that is mapped and prepared for each viewpoint, information table management unit 309 that manages information on the mapping between goal components and extraction results, etc., including user extraction motives and user attributes User information management unit 310 that stores and stores information on orientation, between viewpoints, between start components, between association components, Others are constituted extracted weight setter 311 sets the weights of the results between the like, from the communication unit 312, for transmitting and receiving user terminal 103 via the communication network 102.

Next, functions of the information extraction apparatus 101 will be described in detail.
FIG. 6 is a flowchart showing a processing procedure for outputting the extraction result and the reason why the extraction result is derived in the information extraction apparatus 101 in this embodiment.
After the start of implementation (step S201), the target information specifying unit 301 specifies information to be extracted (step S202).
Next, based on the extraction target information specified by the target information specifying unit 301 and information on user attributes stored and stored in the user information management unit 310, the user information management unit 310 determines the user's orientation. Specify (step S203). Also, reply information input by the user in response to the question transmitted from the information extraction apparatus 101 to the user terminal 103 via the network 102 by transmitting and receiving between the information extraction apparatus 101 and the user terminal 103 by the communication unit 312. , The user's orientation may be specified by, for example, presenting a plurality of user's orientations and having the user select and input them. Note that the user orientation in the present embodiment means the direction in which the user's consciousness to extract information is directed, and more specifically, the user's extraction motivation, the user Refers to the attributes of
Based on the user's orientation, the weighting setting unit 311 sets the weighting that varies for each viewpoint (step S204). However, the weighting is not limited to between the viewpoints, and may be set between the start components, between the associative components, or between the extraction results. For example, the start component may be weighted by asking the user which of the start components is important. Further, the weighting may vary depending on user attributes and the like.
Then, the start component is specified by the start component specifying unit 302 from the extraction target information specified by the target information specifying unit 301 (step S205).
Next, the information table managed by the information table management unit 309 is searched for an information table prepared for each viewpoint by mapping the associative component associated with the start component, in step S205. It is determined whether or not the identified start component is hit (step S206).
If the start component is not hit, the determination result is “No”, the process proceeds to step S211, and the process ends.
On the other hand, when the start component is hit, the associative component associated with the start component is mapped and the associative component mapped to the start component is displayed in the information table prepared for each viewpoint. Extracted by the associative component extraction unit 303. Next, in the information table managed by the information table management unit 309, the goal component associated with the associative component is mapped and mapped to the associated component in the information table prepared for each viewpoint. The goal component extraction unit 304 extracts the goal component that has been set. Further, based on the mapping information of the goal component managed by the information table management unit 309 and the extraction result, the extraction result mapped to the goal component, the mapped start component, associative component, and goal The result and reason extraction unit 305 extracts the result of the series of associations and viewpoints of the constituent elements (step S207). The start component, the associative component, the goal component, and the viewpoint are not limited to all reasons, and some of them may be extracted as the reason. For example, only the associative component, only the associative component and the viewpoint may be extracted as the reason.
Next, the priority of the extraction result and the reason is totalized by the priority totaling unit 306 based on the weighting between the viewpoints, between the start components, between the associative components, or between the extraction results (step S208).
The output rank determination unit 307 determines the output rank of the extraction result and the reason in descending order of the aggregated priority (step S209).
Then, according to the determined output order, the extraction result and the reason why the extraction result has been derived are output by the output unit 308 (step S210), and the process ends (step S211).

In this embodiment, unlike the case of the first embodiment, the information extraction apparatus 101 stores and stores information on user orientation, including user extraction motives and user attributes, and specifies or stores user information. The management unit 310, the weight setting unit 311 in which a predetermined weight is set for each viewpoint, and the priority aggregation unit 306 that aggregates the priority of the extraction result and the reason after reflecting the weight set by the weight setting unit 311 And an output rank determination unit 307 for determining the output rank of the extraction results and reasons in descending order of the priority tabulated by the priority tabulation unit 306, and the output unit 308 outputs the reasons for each extraction result, The reason for outputting for each extraction result is output according to the output order determined by the output order determining unit 307.
The extraction result output by the output unit 308 is output in accordance with the output order determined by the output order determination unit 307.
Further, the priority aggregation by the priority aggregation unit 306 is performed by a kind of point calculation based on the weighting between viewpoints set by the weight setting unit 311. The weighting may be dynamically changed by the user, and may be further changed by the information extraction apparatus 101 learning depending on the usage status of the user.

FIG. 7 shows an output example of the extraction result in the second embodiment and the reason why the extraction result has been derived.
The target of the extraction is “I want to find a restaurant.” The intention of the user is “To date her on a diet. Both of them work in a place where there is a lot of overtime, and they live in different locations but live along the Tokaido Line. Is the case.
In this case, store A and store B are output as extraction results, and the reason why these extraction results have been derived is as follows. Is an example of output from the viewpoint of “location” and “atmosphere”. The reason why the store A is extracted is output from three viewpoints, but in order of “accompanying person”, “location”, and “atmosphere”, reflecting the weighting for each viewpoint. More specifically, in this case, since the purpose of searching for a restaurant is “to date her”, the viewpoint of “accompanied person” is the heaviest. Next, “They work in a workplace with a lot of overtime and live in different locations but live along the Tokaido Line.” The next point of view is “location”.

As another output example, for example, when a certain company wants to create a WEB site but does not know what kind of photograph can be used as a photograph to be posted on the WEB site, the information extraction apparatus 101 is a user concerned. On the other hand, the user's intentions about what kind of object he wants to convey for what purpose is asked, and the photographer's name is output as the extraction result based on the answer. Because the photographer's photograph can logically convey the importance of environmental issues to men and women in their 30s, it is possible to output the photographer's photograph and so on.
Note that the photographer's name may be output as an extraction result, and the photograph taken by the photographer may be output together.

In the conventional search system, the result based on the user's intention is not presented, and the search result is presented based on a judgment criterion completely unrelated to the user's intention. However, in this embodiment, by weighting between viewpoints according to the user's orientation, the extraction result and the priority of the reason can be flexibly changed, and extraction based on the user's orientation Present results and reasons. That is, it is possible to automatically change the viewpoint according to the user's intentionality and output.
According to the present embodiment, the process until obtaining the extraction result is close to human thought, and there is an effect that the user's satisfaction with the output result is further improved. It is possible to present a series of associative flow of thinking up to the extraction result to the user as a story.

The object of the present invention is to supply a software program that implements the functions of the first and second embodiments to a system or apparatus, and the computer of the system or apparatus reads and executes the software program recorded on a storage medium. Is also achieved.

In this case, the program itself read from the storage medium realizes the novel function of the present invention, and the program and the storage medium storing the program constitute the present invention.

As a storage medium for recording a software program, for example, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

Further, after the program read from the storage medium is written to a memory provided in a function expansion board inserted into the computer of the system or apparatus or a function expansion unit connected to the computer, the program This includes the case where the CPU or the like provided in the function expansion board or unit performs part or all of the actual processing based on the above instruction and the above functions are realized by the processing.

DESCRIPTION OF SYMBOLS 101 Information extraction apparatus 102 Communication network 103

User terminal

201, 301 Target information specific |

specification part

202, 302 Start component

specific part

203, 303

Associative component extractor

204, 304

Goal component extractor

205, 305 Result and reason extraction part 306 Priority calculation unit 307 Output

order determination unit

206, 308

Output unit

207, 309 Information table management unit 310 User information management unit 311 Weight setting unit 312 Communication unit

Claims

Start component specifying means for specifying a start component to be extracted;
Mapping means for performing mapping from the start component specified by the start component specifying means to the extraction result;
Reason extraction means for extracting the reason why the extraction result is mapped from the start component by the mapping means;
An information extraction apparatus comprising: an output unit that outputs an extraction result mapped by the mapping unit and a reason extracted by the reason extraction unit.
The mapping means includes
An associative component extracting means for extracting an associative component mapped to the start component specified by the start component specifying means;
Goal component extraction means for extracting a goal component mapped to the association component extracted by the associative component extraction means;
2. The information extracting apparatus according to claim 1, further comprising: a result extracting unit that extracts an extraction result mapped to the goal component extracted by the goal component extracting unit.
The associative component extracting means is an information table in which an associative component associated with a start component is mapped, and the goal component extracting means is in an information table in which a goal component associated with an associative component is mapped. Extraction based on
3. The information extracting apparatus according to claim 2, wherein the output means outputs at least the associative component among a series of associative flows of the mapped start component, associative component, and goal component. .
The information table is prepared for each viewpoint.
The extraction in the associative component extraction means and the goal component extraction means is performed using only information tables belonging to the same viewpoint,
The information extraction apparatus according to claim 3, wherein the reason for the output by the output means includes the viewpoint.
In the information table used for the associative component extracting means, in addition to the mapping of the associative component associated with the start component, other associative components associated with the associative component are mapped. At least one thing is prepared from the same viewpoint,
The associative component extraction unit extracts an associative component mapped to the start component specified by the start component specifying unit, and then maps other associative components associated with the associative component. Extract the mapped associative components using the existing information table,
5. The information extraction according to claim 4, wherein the goal component extraction unit extracts a goal component mapped to the association component finally extracted by the association component extraction unit. apparatus.
A weight setting means for setting a predetermined weight for each viewpoint;
Priority counting means for counting the priority of each reason after reflecting the weight set by the weight setting means;
An output rank determining means for determining an output rank of reasons in descending order of priorities counted by the priority counting means;
6. The output means according to claim 4, wherein the output means outputs a reason for each extraction result, and the reason output for each extraction result is output according to the output order determined by the output order determination means. Information extraction device.
On the computer,
A start component specifying step for specifying a start component to be extracted; and
A mapping step for performing mapping from the start component specified by the start component specifying step to the extraction result;
A reason extraction step for extracting the reason why the extraction result is mapped from the start component by the mapping step;
An information extraction program that realizes an output step of outputting an extraction result mapped by the mapping step and a reason extracted by the reason extraction step.
The mapping step includes
An associative component extraction step of extracting an associative component mapped to the start component specified by the start component specifying step;
A goal component extraction step for extracting a goal component mapped to the associative component extracted by the associative component extraction step;
8. The information extraction program according to claim 7, comprising a result extraction step of extracting an extraction result mapped to the goal component extracted by the goal component extraction step.
The associative component extracting step is an information table in which an associative component associated with a start component is mapped, and the goal component extracting step is in an information table in which a goal component associated with an associative component is mapped. Extraction based on
9. The information extraction program according to claim 8, wherein the output step outputs the result of at least the associative component among a series of associative flows of the mapped start component, associative component, and goal component. .
The information table is prepared for each viewpoint.
The extraction in the associative component extraction step and the goal component extraction step is performed using only information tables belonging to the same viewpoint,
The information extraction program according to claim 9, wherein the reason for outputting in the output step includes the viewpoint.
In the information table used in the associative component extraction step, in addition to the mapping of the associative component associated with the start component, other associative components associated with the associative component are mapped. At least one thing is prepared from the same viewpoint,
The associative component extracting step extracts an associative component mapped to the start component identified by the start component identifying step, and then maps other associative components associated with the associative component. Extract the mapped associative components using the existing information table,
The information extraction according to claim 10, wherein the goal component extraction step extracts a goal component mapped to the association component finally extracted by the association component extraction step. program.
A weight setting step in which a predetermined weight is set for each viewpoint;
A priority counting step of counting the priority of each reason after reflecting the weight set in the weight setting step;
Further realizing an output rank determination step for determining the output rank of the reasons in descending order of priority calculated by the priority calculation step,
12. The output step according to claim 10, wherein the output step outputs a reason for each extraction result, and the reason output for each extraction result is output according to the output order determined by the output order determination step. Information extraction program.