JP2007272485A - Associative retrieval device and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve performance in time of associative retrieval of a life log. <P>SOLUTION: This associative retrieval device has: an inference engine 11 inputting a life log Oi configured from a profile specifying action of a user, and retrieving an associated word of a superior concept associated from a profile of the inputted life log Oi by use of an ontology DB 20 storing ontology in each the profile; a subset retrieval part 12 retrieving the life log Oj including the profile having the retrieved associated word as the superior concept by use of a life log DB 30 and the ontology DB 20; and an output means outputting the retrieved life log Oj. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an associative search device and a computer program.

  In recent years, using a variety of sensor devices (for example, watches, GPS (Global Positioning System), barcode and two-dimensional code readers, digital cameras, etc.) of mobile terminals, users' action history records (life logs) It is studied to generate (see, for example, Non-Patent Document 1). FIG. 12 is a conceptual diagram showing the configuration of the life log. The life log is composed of a profile that identifies the user's behavior. Examples of the profile include an address, time, product, photograph, and text. FIG. 13 shows the types of life logs. For example, a purchase history log is a life log in which a user's purchase history is recorded, and the time when a user purchases a product (time profile), the purchase location of the product (address profile, landmark profile), and the name of the purchased product (product) Profile).

The life log is data obtained mainly by the user, and its characteristic is that the user's preference, action range, etc. are largely biased. The life log is stored in a database server or the like and used for user behavior analysis, self-management, and the like. For example, use for associative search of a life log is mentioned. The life log associative search is a search of a life log database for another life log that a user thinks is associated with a certain life log. By presenting the life log obtained as a result of the associative search to the user, it is possible to realize the user's reflection support for the user's past storage area.
Masaru Honjo, 3 others, “Profile Blog: Realization of Retrieving Lifelog Search Based on Blog”, Proceedings of Multimedia, Distributed, Collaboration and Mobile (DICOMO 2005) Symposium, Information Processing Society of Japan, 2005 July 6, p. 461-464

  However, the above-described conventional technique has a problem that sufficient performance cannot be improved when performing an associative search of a life log.

  The present invention has been made in view of such circumstances, and an object of the present invention is to provide an associative search device capable of improving performance when performing an associative search of a life log.

  Another object of the present invention is to provide a computer program for realizing the associative search device of the present invention using a computer.

  In order to solve the above-described problem, an associative search device according to the present invention includes an input unit that inputs a user behavior history record configured from a profile that identifies a user's behavior, and an association from the input user behavior history record profile. An ontology database that stores an ontology for each profile using an ontology database that stores the ontology for each profile, and a user action history record that includes a profile having the retrieved association word as a superordinate concept. A subset search means for searching using a database and a user action history record database, and an output means for outputting the searched user action history record are provided.

  In the associative search device according to the present invention, the user action history record searched by the subset search means and the input user action history record are weighted based on the ontological distance between profiles using the ontology database. And a degree-of-association calculating means for calculating the degree of association between the two, and outputting the calculated degree of association together with the user action history record searched by the subset searching means.

  In the associative search device according to the present invention, the relevance calculation means calculates the sum of the distances related to all profiles included in the input user action history record.

  In the associative search device according to the present invention, the relevance calculation means has a weighting factor for each profile included in the input user action history record.

  The associative search device according to the present invention is characterized in that a query is issued to the ontology database and a response is obtained.

  The computer program according to the present invention uses an ontology database that accumulates ontologies for each profile, associative words of higher concepts associated with the profile of the first user action history record composed of profiles that specify user actions. And a function of searching for a second user action history record including a profile having the searched associative word as a higher concept using the ontology database and the user action history record database. It is characterized by realizing.

  In the computer program according to the present invention, the first user action history record and the second user action history record are weighted based on an ontology distance between profiles using the ontology database. The computer is further made to realize a function of calculating the degree of association.

  In the computer program according to the present invention, the sum of the distances related to all profiles included in the first user action history record is calculated when the relevance is calculated.

The computer program according to the present invention has a weighting factor for each profile included in the first user action history record, and is used for calculating the relevance.
As a result, the aforementioned associative search device can be realized using a computer.

  ADVANTAGE OF THE INVENTION According to this invention, it becomes possible to aim at the performance improvement at the time of performing an associative search of a life log (user action history record). In addition, since the degree of association between the input life log and the life log of the search result is calculated, the importance of the search result can be easily determined.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing the overall configuration of an associative search system according to an embodiment of the present invention. FIG. 2 is a flowchart of associative search processing according to the present embodiment.

  The associative search system shown in FIG. 1 includes an associative search engine 10, an ontology database (ontology DB) 20, and a life log database (life log DB) 30.

  The associative search engine 10 includes an inference engine 11, a subset search unit 12, a weight assignment unit 13, and an association degree calculation unit 14. The associative search engine 10 receives a life log (user action history record) Oi composed of profiles p1, p2, p3,...

  The inference engine 11 extracts all the profiles p1, p2, p3... Included in the input life log Oi. Then, for each extracted profile, the ontology DB 20 is used to search for an association word of a superordinate concept associated with the profile. Thereby, the associative word is acquired for each of the profiles p1, p2, p3... Included in the life log Oi. Note that associative words are not acquired for search failure profiles.

  The ontology DB 20 stores an ontology (vocabulary space) 200 for each of the profiles p1, p2, p3. Ontology 200 represents a user's knowledge concept. Further, the ontology 200 can be arbitrarily added, edited, and exchanged.

  The ontology 200 has a hierarchical structure with a tree structure, and can be described by, for example, Semantic Web OWL / RDFS. FIG. 3 shows an example of a hierarchical structure for the address profile ontology 200. In FIG. 3, a city name is defined as a superordinate concept one level above the town name of the address profile, and a prefecture name is defined as a superordinate concept one level above. FIG. 4 shows an example of OWL / RDFS description of ontology 200 according to this hierarchical structure.

  In response to the query from the associative search engine 10, the ontology DB 20 acquires, from the ontology 200 of the profile, the word of the high-level concept of the queried profile and returns a response. The returned word is used as a high-level association word associated with the inquiry profile. As described above, the associative search engine 10 is separated from the ontology 200 by accessing the ontology 200 via a query. Thereby, the ontology DB 20 can dynamically add, edit, and exchange the ontology 200 without interrupting the operation of the associative search engine 10.

  For the ontology 200 described in OWL / RDFS, a query can be described using a query language called SPARQL (Simple Protocol And RDF Query Language).

  A plurality of ontology 200 may exist for the same profile. For example, it is assumed that an ontology 200 of an address profile and an ontology 200 of a landmark profile exist, and the ontology 200 of the landmark profile has a hierarchical structure shown in FIG. At this time, the ontology 200 of the address profile and the ontology 200 of the landmark profile can be integrated, and can be integrated as illustrated in FIG. As a result, from the integrated ontology 200, it is possible to acquire, as an associative word, a high-level concept word related to both the address profile and the landmark profile.

  FIG. 7 shows an example of acquired association words for each of the profiles p1, p2, p3... Included in the life log Oi.

  The subset search unit 12 searches the life log Oj including the profile having the associative word acquired by the inference engine 11 as a superordinate concept using the ontology DB 20 and the life log DB 30.

In the search processing, the following processing is performed for each life log with respect to all life logs accumulated in the life log DB 30.
First, all profiles included in the life log are extracted. Next, for each extracted profile, the ontology DB 20 is used to search for an association word of a superordinate concept associated with the profile. Subsequently, the associative word of the search result is collated with the associative word acquired by the inference engine 11. As a result of the collation, when the associative words match, the life log is acquired from the life log DB 30. The acquired life log Oj is output to the weight assignment unit 13 and the relevance calculation unit 14.

  FIG. 8 shows a description example of the life log stored in the life log DB 30. In FIG. 8, the RDF description of the Semantic Web is adopted. In this case, by issuing a SPARQL query illustrated in FIG. 9 to the life log DB 30, all life log subjects (subjects) in the life log DB 30 can be acquired. The result is returned from the query for the number of life logs. According to the query of FIG. 9, region (prefecture name), locality (city name), street (town name), and landmark (landmark name) are returned. For this response, a URL (Uniform Resource Locator) such as “http://www.ubila.kddi.com/adr/Tokyo-Minato-ku-Shibakoen 2-chome-Tokyo Tower” is created. The association word of the superordinate concept is acquired from this URL by inquiring the ontology DB 20 using the SPARQL query illustrated in FIG. The acquired associative word and the associative word acquired by the inference engine 11 are collated, and if they match, the life log is acquired from the life log DB 30.

  The weight assigning unit 13 uses the ontology DB 20 to determine the lifelog Oi input to the associative search engine 10 and the lifelog Oj searched by the subset search unit 12 based on the ontology distance between profiles. Assign weights. In this weight assignment process, for the life log Oi and the life log Oj, an ontology distance between each associative word for a certain profile is obtained using the ontology DB 20.

  FIG. 11 shows an example of a tree structure of ontology 200. The ontology 200 has an upper ontology layer and a lower instance layer. The instance layer corresponds to the value of the profile. The ontological distance between associative words is calculated as the number of branches included in a path connecting two words (associative words) in the instance layer via the ontology layer in the tree structure of ontology 200.

  In the ontology 200 of the profile p1 shown in FIG. 11, the ontology layer has two layers. In the ontology 200 of the profile p1, the path connecting the word “a” and the word “b” passes through one lower layer of the ontology layer, and thus the number of branches included in the path is two. Therefore, the distance between the word “a” and the word “b” is 2. Similarly, the distance between the word “c” and the word “d” is also 2. The distance between the words connected by paths that pass through the upper layers of the ontology layer, for example, the word “a” and the word “c” is 4, since the number of branches included in the path is 4.

As a specific example, the weight calculation will be described using an example of an address profile.
“Tokyo-Minato-ku Shiba Koen 2-Chome-Tokyo Tower” and “Tokyo-Minato-ku Shiba Koen 2-Chome-Tokyo Tower” have exactly the same tree structure data, and the distance at this time is 0 .
"Tokyo-Minato-ku-Shiba Park 2-chome-Tokyo Tower" and "Tokyo-Minato-ku-Shiba Park 2-chome-Minato Hamamatsu Station XXX Station" have the same tree structure above the parent "Shiba Park 2-chome" This is data, and the distance at this time is 2.
“Tokyo-Minato-ku-Shiba-koen 2-chome-Tokyo Tower” and “Tokyo-Minato-ku-Azabu-koen 2-chome-YYY Station” have the same tree structure data for the parent “Minato-ku”. The distance is 4.
“Tokyo-Minato-ku-Shiba-koen 2-chome-Tokyo Tower” and “Tokyo-Shinjuku-ku-Haramachi 1-chome-ZZZ Station” are the same tree structure data of the parent parent “Tokyo” The distance at this time is 6.
“Tokyo-Minato-ku-Shiba Koen 2-chome-Tokyo Tower” and “Hokkaido-Sapporo City Teine-ku-Maeda Gojo 9-chome-WWW Station” have no matching points. In this case, the distance is 8.

  Also, the distance can be calculated in the same manner for other types of profiles. For example, for a product profile, an appropriate distance can be obtained by performing ontology according to a predetermined product classification system.

  The weight for each branch can be arbitrarily set. In the above example, the distance is calculated by setting the weights of all branches to 1, but the weights of the branches may be changed according to the hierarchy.

  The weight assigning unit 13 may assign a weight for each profile. Also, weighting may be determined based on the life log impulse characteristics, palatability, forgetting index, and time density.

  The relevance calculation unit 14 calculates the relevance between the life log Oi input to the associative search engine 10 and the life log Oj searched by the subset search unit 12 using the weight determined by the weight assignment unit 13. To do. The degree of relevance is calculated by, for example, the calculation formula (1).

  However, S (Oi, Oj) is the degree of association between the life log Oi and the life log Oj. Dpl (Oi, Oj) is an ontological distance between the associative words of the life log Oi and the life log Oj for the profile pl. Further, ωpl is a weight for the profile pl.

  In Equation 1, the relevance S (Oi, Oj) between the life log Oi and the life log Oj is calculated as the sum of the distances dpl (Oi, Oj) regarding all the profiles pl included in the life log Oi. Furthermore, weighting for each profile pl is possible. A smaller value of the degree of association indicates that the relationship between the life log Oi and the life log Oj is higher.

  The relevance calculation unit 14 outputs the calculated relevance S (Oi, Oj) and the life log Oj. Thereby, together with the life log Oj associated with the input life log Oi, the degree of association S (Oi, Oj) representing the relation between the two is obtained.

  Note that the output order of the life log Oj may be determined based on the value of the relevance S (Oi, Oj). For example, the life logs Oj may be rearranged and output in ascending order of relevance value, that is, in descending order of relevance.

  Further, a threshold is set in advance for the degree of association S (Oi, Oj), and only the life log Oj having the degree of association S (Oi, Oj) less than or equal to the threshold is output, and selection of the life log Oj to be output is selected. May be performed.

  As described above, the associative search engine 10 according to the present embodiment first acquires the life log Oi (step S1 in FIG. 2). Next, the associative search engine 10 performs inference using an ontology for all profiles pj included in the life log Oi, and acquires a superordinate concept (step S2 in FIG. 2). Next, the associative search engine 10 searches the life log DB 30 for a life log Oj including a profile having the same superordinate concept as the acquired superordinate concept (step S3 in FIG. 2). Next, the associative search engine 10 assigns weights based on the ontological distance for the life log Oi and the life log Oj (step S4 in FIG. 2). Next, the associative search engine 10 calculates the degree of association between the life log Oi and the life log Oj using the weight (step S5 in FIG. 2).

  According to the present embodiment, a life log including a profile having the same associative word as an upper word concept associated with an upper word associative word associated with the profile included in the input life log is converted into a database using the ontology DB and the life log DB. Get by search. Therefore, even if the profile value is a word and is not a numerical value that can be arithmetically calculated, a life log having a profile value that is semantically close to the profile value included in the input life log is associated with the result. Can be given as. This makes it possible to improve the performance when performing an associative search of a life log. In addition, since the degree of association between the input life log and the life log of the search result is calculated, the importance of the search result can be easily determined.

  In general, most of the profile values constituting the life log are words, and the semantic distance between the two cannot be obtained by numerical calculation between the profile values. However, according to this embodiment, even if the value of the profile is a numerical value or a word, the distance between the two can be obtained. This can be expected to greatly improve the performance of associative search of life logs.

  As described above, according to the present embodiment, a life log highly relevant to the input life log can be searched from the user's life log history (life log database). It is possible to effectively perform reflection support and memory landing.

In addition, by recording a program for realizing the function of the associative search engine 10 shown in FIG. 1 on a computer-readable recording medium, causing the computer system to read and execute the program recorded on the recording medium, An associative search process may be performed. Here, the “computer system” may include an OS and hardware such as peripheral devices.
Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.
The “computer-readable recording medium” means a flexible disk, a magneto-optical disk, a ROM, a writable nonvolatile memory such as a flash memory, a portable medium such as a CD-ROM, a hard disk built in a computer system, etc. This is a storage device.

Further, the “computer-readable recording medium” means a volatile memory (for example, DRAM (Dynamic DRAM) in a computer system that becomes a server or a client when a program is transmitted through a network such as the Internet or a communication line such as a telephone line. Random Access Memory)), etc., which hold programs for a certain period of time.
The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.
The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes design changes and the like within a scope not departing from the gist of the present invention.

It is a block diagram which shows the whole structure of the associative search system which concerns on one Embodiment of this invention. It is a flowchart of the associative search process according to the embodiment. It is an example of the hierarchical structure of ontology 200 shown in FIG. It is an example of description of ontology 200 shown in FIG. It is an example of the hierarchical structure of ontology 200 shown in FIG. It is an example of description of ontology 200 shown in FIG. It is an acquisition example of the associative word which concerns on one Embodiment of this invention. It is a description example of the life log stored in the life log DB 30 shown in FIG. It is an example of the query issued to life log DB30 shown in FIG. It is an example of the query issued to ontology DB20 shown in FIG. It is an example of the tree structure of ontology 200 shown in FIG. It is a conceptual diagram which shows the structure of a life log. It is a chart which shows the kind of life log.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Associative search engine (associative search apparatus), 11 ... Inference engine, 12 ... Subset search part, 13 ... Weight allocation part, 14 ... Relevance calculation part, 20 ... Ontology database (ontology DB), 30 ... Life log database ( Life log DB), 200 ... Ontology (vocabulary space), Oi, Oj ... Life log (user action history record), p1, p2, p3 ... Profile, S (Oi, Oj) ... Relationship between life log Oi and life log Oj Every time

Claims (9)

An input means for inputting a user action history record composed of a profile for identifying a user action;
An inference means for searching an association word of a superordinate concept associated with the profile of the input user behavior history record using an ontology database that accumulates the ontology for each profile;
Subset search means for searching a user action history record including a profile having the searched associative word as a superordinate concept using the ontology database and the user action history record database;
Output means for outputting the retrieved user action history record;
An associative search device characterized by comprising: With respect to the user action history record searched by the subset search means and the input user action history record, the ontology database is used to perform weighting based on the distance on the ontology between profiles, and the degree of relevance between them is calculated. A relevance calculation means,
2. The associative search device according to claim 1, wherein the calculated relevance is output together with a user action history record searched by the subset search means.   The associative search device according to claim 2, wherein the relevance calculation means calculates a sum of the distances related to all profiles included in the input user action history record.   4. The associative search device according to claim 2, wherein the relevance calculation means includes a weighting factor for each profile included in the input user action history record.   The associative search device according to any one of claims 1 to 4, wherein a query is issued to the ontology database and a response is obtained. A function of searching for an associative word of a superordinate concept associated with a profile of a first user action history record configured from a profile for identifying a user's action using an ontology database that accumulates ontologies for each profile;
A function of searching for a second user action history record including a profile having the searched associative word as a superordinate concept using the ontology database and the user action history record database;
A computer program for causing a computer to realize the above.   A function of performing weighting based on an ontology distance between profiles using the ontology database and calculating a degree of association between the first user action history record and the second user action history record; The computer program according to claim 6, which is realized by a computer.   The computer program according to claim 7, wherein when calculating the relevance, the sum of the distances related to all profiles included in the first user action history record is calculated. 9. The computer program according to claim 7, wherein the computer program has a weighting factor for each profile included in the first user action history record and is used for calculating the relevance.

Images