JP2017146926A - Information processing device, information processing method and information processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information processing device, an information processing method and an information processing program which can determine ambiguity of a keyword to be used for retrieval with high accuracy.SOLUTION: An information processing device 3 includes a storage part, and an ambiguity determination part. The storage part associates a keyword used during retrieval with information on a plurality of objects selected among retrieval results retrieved by using the keyword to be stored as retrieval history information. The ambiguity determination part calculates a similarity among the plurality of objects corresponding to the keyword on the basis of the retrieval history information stored in the storage part, and determines the ambiguity of the keyword from the calculated similarity.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an information processing apparatus, an information processing method, and an information processing program.

  2. Description of the Related Art Conventionally, there is a service for publishing a website on a network such as the Internet and presenting information desired by a user of the terminal apparatus to a terminal apparatus that has accessed the website.

  For example, in an online shopping website, a search query including a keyword specified by a user is received from a terminal device, a database in which product information is registered in advance is searched based on the keyword, and information on the searched product is included. The search result is transmitted to the terminal device. If the searched product is information on the product intended by the user, the product can be provided to the user by performing a purchase procedure or the like.

  In addition, when a user inputs a search keyword and clicks the “search” button, a mechanism for displaying not only a search result page but also other search keywords related to the search keyword input by the user is proposed. (For example, refer to Patent Document 1).

  In the technique described in Patent Literature 1, a search keyword input by a user is received, and a plurality of predetermined words corresponding to intentions not clearly defined are compared with the search keyword. Then, based on the comparison result, it is determined whether or not the search keyword indicates the intention clearly defined by the user, and if it is determined that the search keyword does not indicate the clearly defined intention, further Processing for acquiring the recommended search keyword is executed.

Special table 2014-501422 gazette

  However, since the above-described conventional method uses an unclear intention word set that is a plurality of predetermined words corresponding to intentions that are not clearly defined, for example, depending on the setting of the unclear intention word set, the ambiguity of the keyword may be reduced. There is a problem in determination accuracy.

  The present invention has been made in view of the above, and an object thereof is to provide an information processing apparatus, an information processing method, and an information processing program capable of accurately determining the ambiguity of a keyword used for a search. .

  The information processing apparatus according to the present application includes a storage unit and an ambiguity determination unit. The storage unit stores, as search history information, a keyword used at the time of search and information on a plurality of targets selected from search results searched using the keyword in association with each other. The ambiguity determination unit calculates the similarity between the plurality of objects corresponding to the keyword based on the search history information stored in the storage unit, and determines the ambiguity of the keyword from the calculated similarity .

  According to one aspect of the embodiment, it is possible to provide an information processing apparatus, an information processing method, and an information processing program that can accurately determine the ambiguity of a keyword used for a search.

FIG. 1 is an explanatory diagram of information processing according to the embodiment. FIG. 2 is a diagram illustrating a configuration example of the information processing system according to the embodiment. FIG. 3 is a diagram illustrating a specific configuration example of the information processing apparatus according to the embodiment. FIG. 4 is a diagram illustrating an example of an ambiguity table stored in the ambiguity information storage unit. FIG. 5 is a diagram illustrating an example of a search history table stored in the search history storage unit. FIG. 6 is a diagram illustrating an example of search result information displayed on the terminal device. FIG. 7 is a diagram illustrating an example of a topic distribution to be selected. FIG. 8 is a diagram illustrating an example of feature information stored in the feature information storage unit. FIG. 9 is a diagram illustrating topic distributions of different queries. FIG. 10 is a flowchart illustrating a flow of information processing in the information processing apparatus according to the embodiment. FIG. 11 is a flowchart showing the flow of the search process shown in FIG. FIG. 12 is a flowchart showing the flow of the ambiguity determination process shown in FIG. FIG. 13 is a diagram illustrating an example of a hardware configuration of a computer that executes an information processing program.

  Hereinafter, a mode for carrying out an information processing apparatus, an information processing method, and an information processing program according to the present application (hereinafter referred to as “embodiment”) will be described in detail with reference to the drawings. Note that the information processing apparatus, the information processing method, and the information processing program according to the present application are not limited by this embodiment.

[1. Information processing including ambiguity determination processing]
First, information processing according to the embodiment will be described. FIG. 1 is an explanatory diagram of information processing according to the embodiment, and the information processing is executed by the information processing apparatus 3. Information processing executed by the information processing apparatus 3 includes search processing and ambiguity determination processing. Hereinafter, the search process and the ambiguity determination process will be described in this order.

  As illustrated in FIG. 1, when a search query including a keyword k1 (hereinafter referred to as a search keyword k1) is transmitted from the terminal device 2 (step S1), the information processing apparatus 3 according to the embodiment receives the search query. A search process is performed using the included search keyword k1, and the search result is transmitted to the terminal device (step S2).

  The search result transmitted from the information processing device 3 includes information on a plurality of selection targets d1 to dn (n is an integer of 2 or more). A user of the terminal device 2 (hereinafter sometimes referred to as a terminal user) selects one or more selection targets d from among the plurality of selection targets d1 to dn (hereinafter may be collectively referred to as selection targets d). When selected, the selection result is transmitted to the information processing apparatus 3 (step S3). When the information processing device 3 receives the selection result from the terminal device 2, the information processing device 3 transmits information corresponding to the selection result to the terminal device 2 (step S4).

  For example, when the information processing device 3 provides a website for online shopping, the information processing device 3 sets the links to the web pages of a plurality of products corresponding to the search keyword k1 as the selection targets d1 to dn. The search result page is transmitted to the terminal device 2. When the terminal user selects the selection target d1 from the search result page (clicks on the link for the product A), the information processing device 3 provides the terminal device 2 with a web page corresponding to the selection target d1 (product A). .

  The information processing apparatus 3 stores the history of the search processing performed in steps S1 to S4 as described above as search history information (step S5). In the example shown in FIG. 1, the selection ratio by the terminal user of each of the selection targets d1 to dn is stored as search history information in association with the search keyword k1. This selection ratio is a ratio with respect to the entire selection result of one or more terminal users for the same search keyword k1.

  Further, the information processing device 3 analyzes the features of the selection targets d1 to dn and stores them as feature information in the storage unit (step S6). The feature information is an index for indicating the feature of the selection target d. For example, when the selection target d is a web page, the information processing apparatus 3 extracts a feature element included in the web page and detects a distribution of the feature element in the web page.

  The feature element is, for example, a topic extracted from a word included in a web page. Such a topic is an abstract representation of a word included in data such as a web page, for example, as a topic or interest. For example, the information processing apparatus 3 can extract topics using, for example, a latent Dirichlet Allocation (LDA) which is a known topic model (language model).

  The information processing device 3 calculates the similarity between the plurality of selection targets d1 to dn corresponding to the search keyword k1, and determines the ambiguity of the search keyword k1 from the calculated similarity (step S7). For example, the information processing device 3 can calculate the similarity between the selection targets d1 to dn based on the similarity of the distribution of the feature elements between the selection targets d1 to dn. In this case, for example, the information processing device 3 performs weighting of the feature elements of the selection targets d1 to dn according to the selection ratio of the selection targets d1 to dn described above, and between the distributions of the feature elements thus weighted. The similarity is calculated.

  Note that the information processing device 3 directly uses the similarity between the plurality of selection targets d1 to dn as the ambiguity of the search keyword k1, or the three levels of ambiguity according to the similarity between the plurality of selection targets d1 to dn. Can be selected from “low”, “medium”, and “high”.

  As described above, the information processing device 3 determines the ambiguity of the search keyword k1 included in the search query from the similarity between the selection targets d1 to dn (an example of the target) selected by the terminal user. It is possible to accurately determine the ambiguity of the search keyword k1 used.

  In addition, two or more selection targets d are grouped and the selection targets d1 to dn are distributed to a plurality of groups g1 to gm (m <n), and between groups g1 to gm (hereinafter, sometimes collectively referred to as a selection group g). It is also possible to determine the ambiguity of the search keyword k1 included in the search query from the similarity. When the information processing device 3 provides an online shopping website, the selection group g is, for example, a product category.

  In this case, the information processing device 3 analyzes the characteristics of each of the groups g1 to gm (an example of the target) and stores them as characteristic information in the storage unit. For example, when the selection targets d1 to dn are web pages, the information processing device 3 extracts feature elements included in the web pages included in the groups g1 to gm, and calculates the distribution of the feature elements in the groups g1 to gm. To detect.

[2. Information processing system]
FIG. 2 is a diagram illustrating a configuration example of the information processing system according to the embodiment. As illustrated in FIG. 2, the information processing system 1 according to the embodiment includes a plurality of terminal devices 2 ₁ to 2 _n (n is an integer of 2 or more) and an information processing device 3.

The plurality of terminal devices 2 _{1 to} 2 _n (hereinafter may be collectively referred to as the terminal device 2) and the information processing device 3 are connected to each other via the network 4 so as to communicate with each other. The network 4 is, for example, a LAN (Local Area Network) or a WAN (Wide Area Network) such as the Internet.

  The terminal device 2 is, for example, a PC (Personal Computer), a PDA (Personal Digital Assistant), a smartphone, or the like used by a terminal user. In the terminal device 2, for example, an application (for example, a browser application) for accessing the information processing device 3 is installed.

  The information processing apparatus 3 includes a communication unit 10, a storage unit 20, and a control unit 30. The control unit 30 includes an information providing unit 31 and an ambiguity determination unit 32. The information providing unit 31 publishes websites such as a shopping site, a news site, a Q & A site, an electronic commerce site, a video providing site, and a search site on the network 4, and based on the provided information stored in the storage unit 20. Thus, the information providing service is provided to the terminal device 2.

  The terminal device 2 transmits a search query including a search keyword designated for accessing desired information by the terminal user to the information processing device 3, and the information processing device 3 performs a search process according to the search keyword. The search result is provided to the terminal device 2. When the terminal user selects a desired target from the search results, the selection result is transmitted from the terminal device 2 to the information processing device 3. The information processing device 3 provides information corresponding to the selection result of the terminal user to the terminal device 2. Note that the search keyword includes one or more words (for example, one word or a set of a plurality of words).

  The storage unit 20 stores the search keyword used at the time of search and the information related to a plurality of objects selected from the search results searched using the search keyword as search history information in association with each other. The ambiguity determination unit 32 calculates the similarity between a plurality of objects corresponding to the search keyword based on the search history information stored in the storage unit 20, and determines the ambiguity of the search keyword from the calculated similarity.

  The information providing unit 31 can provide a search result corresponding to the search keyword based on the ambiguity determined by the ambiguity determining unit 32. For example, when the ambiguity of the search keyword is high, the information providing unit 31 performs a search by supplementing the search query with other information (for example, an additional keyword), or other information for assisting the search (for example, , Additional keywords, etc.) can be provided to the terminal device 2 as narrowing candidate information. The refinement candidate information is displayed on the search screen of the terminal device 2 (for example, at the top of the search keyword input box) so that the terminal user can select it, and the refinement candidate is selected by the terminal user in the terminal device 2. The selection result is transmitted from the terminal device 2 to the information processing device 3. Thereby, the information provision part 31 can provide the terminal user with the search result according to the narrowing-down candidate selected by the terminal user. The narrowing candidate information may be information including a search keyword and an additional keyword. Note that as the other information, for example, a keyword corresponding to preference information can be used using user information of the terminal user.

  In this way, the information processing device 3 calculates the similarity between a plurality of targets (for example, the selection target d or the selection group g) corresponding to the search keyword, and determines the ambiguity of the search keyword from the calculated similarity. Therefore, the ambiguity of the search keyword can be accurately determined. Furthermore, the information processing device 3 presents an appropriate search result to the terminal user based on the ambiguity of the search keyword, or provides the terminal user with information on the narrowing candidate to cause the terminal user to select the narrowing candidate. The search intention of the terminal user can be made clearer.

  Hereinafter, the configuration of the information processing apparatus 3 will be described more specifically with reference to FIG. FIG. 3 is a diagram illustrating a specific configuration example of the information processing apparatus 3 according to the embodiment. Hereinafter, the communication unit 10, the storage unit 20, and the control unit 30 will be described in detail.

[2.1. Communication unit 10]
The communication unit 10 is a communication interface that transmits and receives information to and from the network 4, and connects to the network 4 by wire or wirelessly. The control unit 30 can transmit and receive various types of information to and from the terminal device 2 via the communication unit 10 and the network 4.

[2.2. Storage unit 20]
The storage unit 20 includes a provision information storage unit 21, a user information storage unit 22, an ambiguity information storage unit 23, a search history storage unit 24, and a feature information storage unit 25. The provided information storage unit 21, the user information storage unit 22, the ambiguity information storage unit 23, the search history storage unit 24, and the feature information storage unit 25 are, for example, a semiconductor memory element such as a RAM (Random Access Memory), a flash memory, or the like. , A storage device such as an HDD (Hard Disk Drive) or an optical disk.

[2.3. Control unit 30]
The control unit 30 includes, for example, a microcomputer having a central processing unit (CPU), a read only memory (ROM), a random access memory (RAM), an input / output port, and various circuits.

  The control unit 30 includes an information providing unit 31 and an ambiguity determination unit 32. The functions of the information providing unit 31 and the ambiguity determining unit 32 are realized, for example, by the CPU reading and executing a program stored in the ROM. The information providing unit 31 and the ambiguity determining unit 32 may be partially or entirely configured by hardware such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

  The information provision unit 31 includes a request reception unit 41, a search processing unit 42, a provision information generation unit 43, and a provision information output unit 44. The ambiguity determination unit 32 includes a feature extraction unit 51, an information acquisition unit 52, a similarity calculation unit 53, and a query ambiguity setting unit 54. Hereinafter, each of these parts will be described in detail.

[2.3.1. Request receiving unit 41]
The request receiving unit 41 receives various information provision requests transmitted from the terminal device 2 via the network 4. Examples of the information provision request include a search query including a search keyword and an information provision query including information for specifying requested information (for example, URL (Uniform Resource Locator)).

[2.3.2. Search processing unit 42]
The search processing unit 42 acquires a search keyword included in the search query received by the request receiving unit 41 and refers to search target data (for example, product data) stored in advance in the provided information storage unit 21. Search for a plurality of contents corresponding to the search keyword.

  When the information providing unit 31 provides a shopping site service, the content is, for example, a product web page (hereinafter, referred to as a product page) corresponding to the search keyword, and the search processing unit 42 includes the product page. (For example, product name and URL information) is acquired from the provided information storage unit 21. The provided information storage unit 21 is, for example, a database configured to be able to search for search target data.

  In addition, when the information providing unit 31 provides a service of a video providing site, the content is, for example, video data corresponding to a search keyword, and the search processing unit 42 includes information on the video data (for example, a video title) And URL information).

  Further, the search processing unit 42 can change the content of the search processing based on the ambiguity table stored in the ambiguity information storage unit 23. FIG. 4 is a diagram illustrating an example of an ambiguity table stored in the ambiguity information storage unit 23.

  As shown in FIG. 4, in the ambiguity table, the ambiguities of the search keywords k1 to kp (p is an integer of 2 or more) are collectively referred to as the search keywords k1 to kp (hereinafter collectively referred to as the search keyword k). Information). In the ambiguity table shown in FIG. 4, “1.21” is associated with the search keyword k1 as ambiguity, “0.88” is associated with the search keyword k2 as ambiguity, and the ambiguity is associated with the search keyword kp. “1.82” is associated.

  The search processing unit 42 determines, for example, the ambiguity of the search keyword k based on the ambiguity table, and uses the auxiliary information when the ambiguity is high (for example, equal to or higher than a preset threshold value TH). Search processing can be performed.

  For example, when the ambiguity of the search keyword is high, the search processing unit 42 uses the user information of the terminal user corresponding to the search query among the user information of the terminal user stored in the user information storage unit 22 as preference information. The corresponding keyword is added to the search keyword as auxiliary information and search processing is performed. On the other hand, when the ambiguity of the search keyword is low (for example, less than a preset threshold value TH), the search processing unit 42 performs the search process without supplementing the search keyword with an additional keyword.

  In addition, when the degree of ambiguity is high, for example, the search processing unit 42 uses other information (for example, additional keywords) supplemented to the search keyword k and information on narrowing candidates that can be selected by the terminal user as auxiliary information. Can be provided to the terminal device 2 as follows. The terminal user can obtain a search result corresponding to the narrowing-down candidate by selecting a narrowing-down candidate, and can obtain a search result more according to the search intention. Note that the search processing unit 42 can also switch the usage method of the auxiliary information described above according to the ambiguity of the search keyword k.

  The search processing unit 42 can also change the search method according to the ambiguity of the search keyword k included in the search query. For example, the search processing unit 42 sets a search method for a search query having a search keyword k with a high degree of ambiguity and a search method for a search query having a keyword k with a low degree of ambiguity to different search methods. The search method can be switched according to the ambiguity of k. Further, the search processing unit 42 changes the ranking method of the searched content (for example, web page) and changes the display order of the content in the search result according to the ambiguity of the search keyword k included in the search query. You can also

[2.3.3. Provided information generation unit 43]
The provided information generation unit 43 acquires the result of the search process performed by the search processing unit 42, and generates information including the search result (hereinafter referred to as search result information) as provided information.

  For example, when the information providing unit 31 provides a shopping site service, the search result information is displayed with names of a plurality of products corresponding to the search keyword, and a plurality of product pages corresponding to the search keyword are linked. It is a web page set as.

  Further, when the information providing unit 31 provides a service of a video providing site, the search result information includes, for example, titles of a plurality of video data corresponding to the search keyword, and a plurality of video data corresponding to the search keyword. Is a web page set as a link destination.

  Further, the provision information generation unit 43 acquires content corresponding to information (for example, URL) included in the information provision query received by the request reception unit 41 from the information provision unit 31, and generates provision information including the acquired content. To do.

  Here, it is assumed that the information providing query received by the request receiving unit 41 is a query for requesting transmission of information on the selection target d selected by the terminal user among the selection targets d included in the search result information. In this case, the provided information generation unit 43 stores the keyword corresponding to the search result information and the information related to the selected selection target d in the search history storage unit 24 in association with each other.

  The information regarding the selection target d includes, for example, information for specifying the selection target d selected by the terminal user, information on a ratio (hereinafter referred to as a selection ratio) of selecting each selection target d by the terminal user, and the like. It is.

  FIG. 5 is a diagram illustrating an example of a search history table stored in the search history storage unit 24. In the search history table (an example of search history information) illustrated in FIG. 5, for example, the search keyword k1 and information on the selection ratio of the selection targets d1 to dn are stored in association with each other.

  In the search history table shown in FIG. 5, the selection ratios of the selection targets d1, d2, d3, d4,..., Dn included in the search result by the search keyword k1 are 5%, 3%, 12%, 2 %,..., 4%. Hereinafter, the distribution of the selection ratio by the terminal user of the selection target d included in the search result by a certain search keyword k may be described as a click distribution p (d | q). Note that the search history information stored in the search history storage unit 24 is not limited to the search history table illustrated in FIG. 5, and is, for example, a table in which the search keyword k and the selection ratio of each selection group g are associated. Hereinafter, the distribution of the selection ratio of the selection group g may be referred to as a click distribution p (g | q).

[2.3.4. Provided information output unit 44]
The provision information output unit 44 transmits the provision information generated by the provision information generation unit 43 to the terminal device 2 via the communication unit 10 that has transmitted a corresponding information provision request (for example, a search query or an information provision query). Thereby, in the terminal device 2, the information according to the information provision request is acquired and displayed.

  FIG. 6 is a diagram illustrating an example of search result information displayed on the terminal device 2. The search result information illustrated in FIG. 6 indicates a product search result page 80 displayed on the display unit of the terminal device 2 when the information providing unit 31 provides a shopping site service.

  The product search result page 80 shown in FIG. 6 shows a search result when a search is performed by clicking (selecting) the search button 82 in a state where “camera” is input to the input box 81 by the terminal user as a search keyword. . In the product search result page 80, a plurality of areas 83 to 85 in which information of products A, B, and C are displayed so as to be selectable are set as search results.

  For example, when the terminal user selects (for example, clicks) the characters “product A”, the product A is selected as the selection target d, and an information provision query including the selection result is sent from the terminal device 2 to the information processing device 3. Sent to. In addition, for example, when the terminal user selects (for example, clicks) the letter “product B”, the product B is selected as the selection target d, and an information provision query including the selection result is processed from the terminal device 2. It is transmitted to the device 3.

[2.3.5. Feature extraction unit 51]
The feature extraction unit 51 extracts the feature of the selection target d included in the search result corresponding to each search keyword k based on the information stored in the search history storage unit 24 and the information stored in the provided information storage unit 21. To do.

  When the selection target d is a web page, the feature extraction unit 51 extracts a plurality of morphemes by morphological analysis from natural language data included in the web page, and extracts feature words (for example, nouns) from the morphemes. To do. Then, the feature extraction unit 51 determines a topic t (an example of a feature element) to which each extracted feature word belongs using a known topic model (for example, the above-described LDA), and the probability distribution of the topic t in the selection target d. The topic distribution p (t | d) is obtained as follows.

  For example, when the selection target d is moving image data or a web page including moving image data, the feature extraction unit 51 selects feature elements such as feature images, feature sounds, and feature words included in the moving image data as topics. Extracted as t, the distribution of the appearance ratio of each topic t can be obtained.

  Note that the topic distribution p (t | d) is a probability distribution targeting one selection target d, but the feature extraction unit 51 is a probability distribution targeting a selection group g including two or more selection targets d. It is also possible to obtain a topic distribution p (t | g). By grouping in this way, even if there are many selection targets d, it is possible to reduce the computation load of the topic distribution.

  FIG. 7 is a diagram illustrating an example of a topic distribution p (t | d) of a certain selection target d. In the example illustrated in FIG. 7, the ratio of the number of topics t1 to tn to the number of all topics t1 to tn (n is an integer of 2 or more) included in a selection target d is represented as a topic distribution p (t | d). Has been.

  The feature extraction unit 51 stores, as feature information indicating the characteristics of the extracted selection target d, information that associates the information of the topic distribution p (t | d) of each selection target d with the information of each selection target d. Store in unit 25. FIG. 8 is a diagram illustrating an example of feature information stored in the feature information storage unit 25. Note that the feature information is not limited to the feature information illustrated in FIG. For example, the feature information may be information in which information on the topic distribution p (t | g) of each selection group g is associated with information on each selection group g.

  The feature extraction unit 51 can also calculate, for example, feature information representing the feature of the selection target d by a vector on a space vector from the selection target d and store the feature information in the feature information storage unit 25. In the feature information, the feature of the selection target d is expressed by the direction and size of the vector. For generating the vector representation, for example, a technique such as latent semantic indexing (LSI) can be used. Note that the feature extraction unit 51 can also calculate feature information represented by a vector of the selection group g from a plurality of selection targets d included in the selection group g.

[2.3.6. Information acquisition unit 52]
The information acquisition unit 52 acquires search history information stored in the search history storage unit 24. When the search history table shown in FIG. 5 is stored in the search history storage unit 24, the information acquisition unit 52, for example, displays information on the selection ratios of the selection targets d1 to dn corresponding to the search keyword k1 as the search history storage unit 24. Get from.

  The information acquisition unit 52 acquires feature information stored in the feature information storage unit 25. For example, the information acquisition unit 52 acquires the topic distribution p (t | d) of the selection targets d1 to dn corresponding to the search keyword k1 from the feature information storage unit 25. The information acquisition unit 52 can also acquire the topic distribution p (t | g) of each of the plurality of selection groups g corresponding to the search keyword k1 from the feature information storage unit 25.

[2.3.7. Similarity calculation unit 53]
The similarity calculation unit 53 includes information on a plurality of selection targets d corresponding to each search keyword k acquired by the information acquisition unit 52 and each topic distribution p (t | d) of the plurality of selection targets d. Based on this, the similarity between a plurality of selection targets d corresponding to each search keyword k is calculated.

The similarity calculation unit 53 quantifies the topic distribution p (t | d) of the selection target d selected from the search results corresponding to the search keyword k in order to quantify the intensity of the distribution of the topic t. After weighting according to the selection ratio, the average topic distribution p (t | q) is obtained by adding together. For example, the similarity calculation unit 53 obtains an average topic distribution p (t | q) for each search keyword k using the following formula (1).

  The average topic distribution p (t | q) indicates the probability that a certain topic t appears when a search query q having a certain search keyword k is given. In the above formula (1), “q” indicates a search query q having the search keyword k, and the search query q and the search keyword k correspond one-to-one. Therefore, “q” substantially indicates the search keyword k.

  In the above formula (1), “d” indicates the selection target d that has been selected (clicked), and “Dq” indicates a set having the selection target d related to the search query q as an element. “P (d | q)” is the click distribution (selection distribution) described above, and indicates a probability that a certain selection target d is selected (clicked) when a certain search query q is given.

  Next, the similarity calculation unit 53 calculates the similarity between the selection targets d for each search query q. For example, the similarity calculation unit 53 calculates the similarity between the query distribution of each selection target d and the average topic distribution p (t | q) for each search query q, and the average topic distribution p (t | q) The similarity between the selection targets d used as a reference is obtained.

  Further, the similarity calculation unit 53 determines the ambiguity of the search keyword k for each search query q (hereinafter referred to as keyword ambiguity) based on the similarity between the selection targets d. For example, the similarity calculation unit 53 calculates a value obtained by averaging the similarities of all selection targets d, and determines the calculation result as a keyword ambiguity.

Specifically, the similarity calculation unit 53 calculates the similarity between the selection targets d as a Kullback-Leibler divergence using the following formula (2), and further selects the selection target. The query ambiguity amb (q, Dq) is obtained as the keyword ambiguity based on the similarity between d. The query ambiguity amb (q, Dq) is obtained by weighting the amount of information of the Cullback liberer between each topic distribution p (t | d) and the average topic distribution p (t | q) by the selection ratio of the selection target d. Is the averaged value.

  The query ambiguity amb (q, Dq) is not negative, and is zero in the case of the search query q having the search keyword k whose terminal user's search intention is completely clear, and the terminal user's search intention is more ambiguous. The search query q having the search keyword k is a larger value. In the above formula (2), “T” represents a set having topic t as an element, and “Dq” represents a set having selection target d related to search query q as an element as described above. Further, as described above, the search query q and the search keyword k have a one-to-one correspondence. Therefore, the query ambiguity amb (q, Dq) indicates the ambiguity of the keyword.

  Thus, the similarity calculation unit 53 quantifies the ambiguity (keyword ambiguity) of the search query q by comparing the topic distribution p (t | d) between the selection targets d for each search query q. Can be FIG. 9 is a diagram illustrating topic distributions p (t | q) of different search queries q1, q2, and q3. The topic distribution p (t | q) shown in FIG. 9 includes three topics t (topics t1 to t3) and three selection targets d (selection targets d1 to d3) for easy understanding. Here is an example.

  As shown in FIG. 9, the distances between the selection targets d1 to d3 in the topic space composed of the topics t1 to t3 are different, and the search queries q1, the search queries q2, and the search queries q3 are in the order of the selection targets d1 to d3. The distance gets closer. Therefore, it can be understood that the ambiguity (keyword ambiguity) of the search query q is: search query q1> search query q2> search query q3.

  Therefore, as described above, the ambiguity of the search keyword k of each search query q can be accurately determined by comparing the topic distribution p (t | q) between the selection targets d for each search query q. I understand that I can do it.

By the way, when each selection target d has only one topic t and the topic t of each selection target d is independent, the above formula (2) is simplified as shown in the following formula (3). Can do.

  In addition, the similarity calculation unit 53 can obtain the keyword ambiguity by comparing the topic distribution p (t | g) between the selected groups g for each search query q. In this case, the similarity calculation unit 53 can obtain the keyword ambiguity by replacing the selection target d with the selection group g and performing the same processing as described above. For example, the similarity calculation unit 53 replaces the topic distribution p (t | d) with the topic distribution p (t | g) and replaces the click distribution p (d | q) in (1), (2), and the like described above. The keyword ambiguity can be obtained by replacing with the click distribution p (g | q).

  Note that the method for calculating the similarity between the selection targets d (or between the selection groups g) is not limited to the above-described example, and it is sufficient that the similarity between the selection targets d (or between the selection groups g) can be calculated. . The similarity calculation unit 53 calculates the similarity between the reference topic distribution prepared in advance and the topic distribution p (t | d) of each selection target d, and selects the selection target d based on the similarity between the similarities. The similarity between them can also be calculated. Similarly, for the selected group g, the similarity calculation unit 53 calculates the similarity between the reference topic distribution prepared in advance and the topic distribution p (t | g) of each selected group g. It is also possible to determine the similarity.

  Further, for example, when the feature of the selection target d is represented by a vector on a space vector, the similarity calculation unit 53 calculates the similarity between the selection targets d of the vector so that the selection target d Similarity can be calculated. Similarly, for the selected group g, the similarity calculation unit 53 can determine the similarity between the selected groups g of vectors.

[2.3.8. Query ambiguity setting unit 54]
The query ambiguity setting unit 54 sets the query ambiguity amb (q, Dq) as it is in the ambiguity information storage unit 23 as the keyword ambiguity. Further, the query ambiguity setting unit 54 does not use the query ambiguity amb (q, Dq) as the keyword ambiguity as it is, but uses the predetermined conversion table to change the query ambiguity amb (q, Dq) into the keyword ambiguity. It can also be converted.

  The conversion table is, for example, a table in which the keyword ambiguity divided into three stages of “low”, “medium”, and “high” is associated with the value range of the query ambiguity amb (q, Dq). Based on the table, the keyword ambiguity is determined from “low”, “medium”, and “high”.

  The keyword ambiguity for each search keyword k is increased so that the lower the similarity between the selection targets d (or between selected groups g) calculated based on the similarity calculation unit 53 is, the higher the keyword ambiguity is. If it can set to the memory | storage part 23, it will not be limited to the said method.

[3. Processing flow of information processing apparatus 3]
Next, the procedure of information processing in the information processing apparatus 3 will be described with reference to FIG. FIG. 10 is a flowchart showing a flow of information processing in the information processing apparatus 3 according to the embodiment, and the information processing is repeatedly executed.

  As illustrated in FIG. 10, the information providing unit 31 of the information processing device 3 determines whether a search query from the terminal device 2 has been received (step S10). When it determines with having received the search query (step S10; Yes), the information provision part 31 performs a search process (step S11). Such search processing is processing shown in steps S20 to S25 shown in FIG. 11, and will be described in detail later.

  When the process of step S11 is completed, or when it is determined in step S10 that a search query has not been received (step S10; No), it is determined whether an information providing query has been received (step S12). When it determines with having received the information provision query (step S12; Yes), the information provision part 31 transmits the provision information according to an information provision query to the terminal device 2 (step S13), and was requested | required by the information provision query. Information related to the provided information is stored as search history information (step S14). The information related to the provided information is, for example, information related to the selected selection target.

  When the process of step S14 is completed, or when it is determined in step S12 that an information providing query has not been received (step S12; No), the ambiguity determination unit 32 determines whether or not an ambiguity determination trigger has occurred. Determination is made (step S15). The ambiguity determination opportunity is an opportunity that occurs at a predetermined cycle (for example, every day), for example.

  When it is determined that the ambiguity determination is triggered (step S15; Yes), the ambiguity determination unit 32 executes an ambiguity determination process (step S16). Such ambiguity determination processing is processing shown in steps S30 to S33 shown in FIG. 12, and will be described in detail later. When the process of step S16 is completed, or when it is determined in step S15 that there is no ambiguity determination trigger (step S15; No), the control unit 30 ends the process shown in FIG.

  Next, the search process shown in step S11 will be described. FIG. 11 is a flowchart showing the flow of the search process in step S11 shown in FIG.

  As shown in FIG. 11, the information providing unit 31 acquires keyword ambiguity information of the received search query from the storage unit 20 (step S20), and determines whether or not the acquired keyword ambiguity is greater than or equal to a threshold value TH. Determination is made (step S21).

  If it is determined that the keyword ambiguity is equal to or higher than the threshold value TH (step S21; Yes), the information providing unit 31 acquires auxiliary information (step S23). The auxiliary information is information (for example, additional keywords) supplemented to the search keyword included in the search query, for example. When step S23 is completed or when it is determined that the keyword ambiguity is not equal to or higher than the threshold TH (step S21; No), the information providing unit 31 performs a search using the search keyword included in the search query (step S21). S22).

  If the information providing unit 31 has acquired auxiliary information in step S23, the information providing unit 31 uses the auxiliary information in step S22. For example, the information providing unit 31 can perform a search process using a search keyword in which auxiliary information (for example, an additional keyword) is supplemented to the search keyword k. In addition, the information providing unit 31 transmits information on narrowing candidates (for example, additional keywords) to be supplemented to the search keyword k included in the search query to the terminal device 2 as auxiliary information to provide narrowing candidate information to the terminal user. . And the information provision part 31 will perform a search by the keyword which supplemented the refinement candidate to the search keyword k, if the query containing the refinement candidate selected by the terminal user is received.

  When the processes of steps S22 and S23 are completed, the information providing unit 31 transmits information on the search result to the terminal device 2 that is the transmission source of the search query (step S24). And the information provision part 31 memorize | stores the information (for example, the information of the search keyword k or the selection object d) of this search history in the memory | storage part 20 (step S25), and complete | finishes the process shown in FIG.

  Next, the ambiguity determination process shown in step S16 will be described. FIG. 12 is a flowchart showing the ambiguity determination process in step S16 shown in FIG.

  As shown in FIG. 12, the ambiguity determination unit 32 extracts the characteristics of the selection target d (or selection group g) corresponding to each search keyword k (step S30). Next, the ambiguity determination unit 32 acquires information on the selection ratio of the selection target d (or selection group g) (step S31).

  Next, the ambiguity determination unit 32 calculates the similarity between the selection target d or the selection group g based on the information acquired in step S31 (step S32). The ambiguity determination unit 32 weights the feature elements of each selection target d (or selection group g) according to the selection ratio of the selection target d (or selection group g), for example, and the weighted feature By calculating the similarity between the distributions of elements, the similarity between the selection targets d (or selection group g) is calculated.

  Next, the ambiguity determination unit 32 stores the query ambiguity according to the similarity calculated in step S32 in the storage unit 20 (step S33), and ends the process illustrated in FIG. In such processing, for example, the ambiguity determination unit 32 stores the similarity calculated in step S32 as it is in the storage unit 20 as a query ambiguity, or uses the predetermined conversion table to calculate the similarity calculated in step S32. It can also be converted to query ambiguity.

[4. Modified example]
In the above-described embodiment, the example in which the search processing in consideration of the keyword ambiguity and the ambiguity determination processing for determining the keyword ambiguity are executed by the same information processing device 3 has been described. The information processing apparatus that performs the ambiguity determination process may be performed by a separate apparatus. In this case, keyword ambiguity information is transmitted and received between these devices.

  Further, the content provided by the information processing apparatus 3 corresponding to the search keyword k is not limited to web content such as a web page, and may be digital content searched using the search keyword k. The digital content includes, for example, document data, moving image data, audio data, and the like.

  In the embodiment described above, the ambiguity determination unit 32 uses the topic model (for example, LDA) as the feature of the selection target d to determine the similarity between the selection targets d based on the distribution of the topic t. Although shown, it is not limited to such an example. For example, the ambiguity determination unit 32 can also determine the similarity between the selection targets d based on the similarity of the distribution of feature words (for example, preset nouns) between the selection targets d.

  In addition to the selection ratio of the selection target d (or selection group g), the ambiguity determination unit 32 can further perform weighting according to the time taken to select the selection target d (or selection group g). Thereby, the ambiguity of the search keyword k can be determined with high accuracy. In this case, the information providing unit 31 stores information on the time taken until selection of the selection target d (or selection group g) in the storage unit 20 as search history information.

[5. Hardware configuration)
Note that the information processing apparatus 3 in the above-described embodiment is realized by a computer 100 having a configuration as illustrated in FIG. 13 executing an information processing program, for example. FIG. 13 is a diagram illustrating an example of a hardware configuration of a computer that executes an information processing program. The computer 100 includes a CPU 101, a RAM 102, a ROM 103, an HDD 104, a communication interface (I / F) 105, an input / output interface (I / F) 106, and a media interface (I / F) 107.

  The CPU 101 operates based on a program stored in the ROM 103 or the HDD 104 and controls each unit. The ROM 103 stores a boot program executed by the CPU 101 when the computer 100 is started up, a program depending on the hardware of the computer 100, and the like.

  The HDD 104 stores data used by programs executed by the CPU 101. For example, the HDD 104 stores the same data as the provided information storage unit 21, the user information storage unit 22, the ambiguity information storage unit 23, the search history storage unit 24, and the feature information storage unit 25 described in FIG. 3. The communication interface 105 corresponds to the communication unit 10, receives data from other devices via the network 4, sends the data to the CPU 101, and transmits data generated by the CPU 101 to other devices via the network 4.

  The CPU 101 controls an output device such as a display and a printer and an input device such as a keyboard and a mouse via the input / output interface 106. The CPU 101 acquires data from the input device via the input / output interface 106. Further, the CPU 101 outputs the generated data to the output device via the input / output interface 106.

  The media interface 107 reads a program or data stored in the recording medium 108 and provides it to the CPU 101 via the RAM 102. The CPU 101 loads the program from the recording medium 108 onto the RAM 102 via the media interface 107, and executes the loaded program. The recording medium 108 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. Etc.

  When the computer 100 functions as the information processing apparatus 3 according to the above-described embodiment, the CPU 101 of the computer 100 executes the program loaded on the RAM 102 to thereby execute the information providing unit 31 and the ambiguity determination illustrated in FIG. Each function of the unit 32 is realized. The CPU 101 of the computer 100 reads the information processing program from the recording medium 108 and executes the information processing program. However, as another example, the program may be acquired from another device via the network 4.

[6. effect]
The information processing apparatus 3 according to the embodiment includes a storage unit 20 and an ambiguity determination unit 32. The storage unit 20 relates to a search keyword k used at the time of search and a plurality of selection targets d (or a plurality of selection groups g; examples of targets) selected from the search results searched using the search keyword k. The information is stored in association with the search history information. Based on the search history information stored in the storage unit 20, the ambiguity determination unit 32 calculates a similarity between a plurality of selection targets d (or between a plurality of selection groups g) corresponding to the search keyword k. The ambiguity of the search keyword k is determined from the similarity. In this way, the ambiguity of the search keyword k is determined in consideration of the relationship between the selection targets d1 to dn (or between a plurality of selection groups g) selected by the terminal user, so that the search keyword k used for the search is used. Can be accurately determined.

  The ambiguity determination unit 32 includes an information acquisition unit 52 and a similarity calculation unit 53. The information acquisition unit 52 acquires feature information indicating the features of each of the plurality of selected selection targets d (or a plurality of selection groups g). In addition, the similarity calculation unit 53 calculates feature information between a plurality of selection targets d (or between a plurality of selection groups g) weighted based on the selection ratio of the selection target d (or the selection ratio of the selection group g). Based on the similarity, a similarity (hereinafter referred to as a target similarity) between a plurality of selection targets d (or between a plurality of selection groups g) is calculated. Since the target similarity is calculated based on the feature information as described above, the similarity can be calculated with high accuracy.

  In addition, the information acquisition unit 52 acquires information indicating the topic distribution p (t | d) (or p (t | g)) (an example of the distribution of feature elements) as feature information. The similarity calculation unit 53 is based on the similarity of the topic distribution p (t | d) (or p (t | g)) between a plurality of selected selection targets d (or between a plurality of selection groups g). The similarity of the target similarity is calculated. As described above, by using the similarity of the topic distribution p (t | d) (or p (t | g)), the target similarity can be calculated with higher accuracy.

  The selected selection target d is content (or a group of content) selected from the search results, and the feature information is a topic distribution included in the content (or topics included in two or more contents of the group). Distribution). If there is a well-adjusted group in advance, the target similarity is calculated based on the topic distribution of the selected group compared to the target similarity calculated based on the topic distribution of the content. There is a possibility that the similarity can be calculated more accurately. Further, by calculating the target similarity based on the topic distribution of the selected group, for example, the determination target can be reduced because the determination target can be reduced, for example.

  The ambiguity determination unit 32 stores the ambiguity information including the keyword ambiguity corresponding to the similarity for each search keyword k in the storage unit 20. As described above, since the keyword ambiguity corresponding to the similarity is stored for each search keyword k, the search according to the keyword ambiguity can be performed on the search keyword k input for performing a new search. it can.

  The information processing apparatus 3 according to the embodiment includes an information providing unit 31. The information providing unit 31 acquires a keyword ambiguity (an example of ambiguity information) for the search keyword k from the ambiguity information for the search keyword k input for a new search, and acquires the acquired keyword ambiguity. Information provision processing (for example, search processing) corresponding to the degree information is performed. Thereby, an appropriate search result according to the ambiguity of the search keyword k of the search query q can be provided to the terminal user.

  The information providing unit 31 compares the keyword ambiguity with a preset threshold value TH, and when the keyword ambiguity is equal to or greater than the threshold value TH, the search keyword k is supplemented with other information, and the search is performed with the supplemented content. If the keyword ambiguity is not greater than or equal to the threshold value TH, the search is performed with the search keyword k. Thereby, an appropriate search result corresponding to the user's search intention can be provided to the terminal user for the search query having the search keyword k having a high degree of ambiguity.

  The information providing unit 31 compares the keyword ambiguity with a preset threshold value TH, and if the keyword ambiguity is equal to or greater than the threshold value TH, information on narrowing candidates for supplementing the search keyword k with other information is displayed on the terminal device. 2 can also be provided. As a result, for a search query having a search keyword k with a high degree of ambiguity, it is possible to provide a terminal user with a clearer infiltration candidate that corresponds to the user's search intention.

  As described above, some of the embodiments of the present application have been described in detail based on the drawings. It is possible to implement the present invention in other forms with improvements.

  Further, the information processing apparatus 3 described above may be realized by a plurality of server computers, and depending on functions, an external platform or the like may be realized by calling an API (Application Programming Interface) or network computing. Can be changed flexibly.

  Moreover, the above-mentioned “section (module, unit)” can be read as “means”, “circuit”, and the like. For example, the ambiguity determination unit can be read as ambiguity determination means or an ambiguity determination circuit.

DESCRIPTION OF SYMBOLS 1 Information processing system 2 Terminal device 3 Information processing apparatus 4 Network 10 Communication part 20 Storage part 21 Provided information storage part 22 User information storage part 23 Ambiguity information storage part 24 Search history storage part 25 Feature information storage part 30 Control part 31 Information Provision unit 32 Ambiguity determination unit 41 Request reception unit 42 Search processing unit 43 Provision information generation unit 44 Provision information output unit 51 Feature extraction unit 52 Information acquisition unit 53 Similarity calculation unit 54 Query ambiguity setting unit

Claims (10)

A storage unit that associates and stores, as search history information, a keyword used at the time of search and information related to a plurality of targets selected from search results searched using the keyword;
Based on the search history information stored in the storage unit, the similarity between the plurality of objects corresponding to the keyword is calculated, and an ambiguity determination unit that determines the ambiguity of the keyword from the calculated similarity,
An information processing apparatus comprising: The ambiguity determination unit
An information acquisition unit that acquires characteristic information indicating characteristics of each of the plurality of objects;
A similarity calculation unit that calculates the similarity between the plurality of objects based on the similarity of the feature information between the plurality of objects weighted based on the selection ratio of the objects. The information processing apparatus according to claim 1. The information acquisition unit
Obtaining information indicating the distribution of feature elements as the feature information,
The similarity calculation unit includes:
The information processing apparatus according to claim 2, wherein the similarity between the plurality of objects is calculated based on the similarity of the distribution of the feature elements between the plurality of objects. The target is a content selected from the search results or a group of the content,
The information processing apparatus according to claim 3, wherein the feature information is a distribution of feature elements included in the content or a distribution of feature elements included in two or more contents of the group. The ambiguity determination unit
The information processing apparatus according to any one of claims 1 to 4, wherein ambiguity information corresponding to the similarity is stored in the storage unit for each keyword. An information providing unit that obtains ambiguity information for the keyword from the ambiguity information and performs an information provision process corresponding to the obtained ambiguity information with respect to the keyword newly entered for performing a search. The information processing apparatus according to claim 5, further comprising: The information providing unit
The ambiguity is compared with a preset threshold, and when the ambiguity is equal to or greater than the threshold, the keyword is supplemented with other information, and a search is performed with the supplemented content. The ambiguity is the threshold The information processing apparatus according to claim 6, wherein, if not, the search is performed using the keyword. The information providing unit
The ambiguity is compared with a preset threshold value, and when the ambiguity is equal to or greater than the threshold value, other information supplemented to the keyword is provided to the terminal device that transmitted the keyword. The information processing apparatus according to claim 6 or 7. An information processing method executed by a computer,
A storage step of associating and storing in a storage unit as search history information the keywords used during the search and the information related to a plurality of objects selected from the search results searched using the keywords;
Based on the search history information stored in the storage unit, the similarity between the plurality of objects corresponding to the keyword is calculated, and the ambiguity determination step of determining the ambiguity of the keyword from the calculated similarity. Information processing method characterized by including. A storage procedure for associating a keyword used at the time of search with information related to a plurality of objects selected from search results searched using the keyword in the storage unit as search history information;
Based on the search history information stored in the storage unit, the similarity between the plurality of objects corresponding to the keyword is calculated, and the ambiguity determination procedure for determining the ambiguity of the keyword from the calculated similarity,
An information processing program for causing a computer to execute.

Images