JP2017045196A - Ambiguity evaluation device, ambiguity evaluation method, and ambiguity evaluation program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To quantify the ambiguity of user's intention so as to present an appropriate retrieval result to a user.SOLUTION: An ambiguity evaluation device includes: storage means that stores a keyword used for retrieval and categories corresponding to retrieval results obtained by retrieval from retrieval target data by using the keyword, in association with each other; and ambiguity evaluation means that calculates the similarity between the categories for each keyword, and evaluates the ambiguity of the keyword from the calculated similarity.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an ambiguity evaluation apparatus, an ambiguity evaluation method, and an ambiguity evaluation program for evaluating ambiguity with respect to a keyword used at the time of search.

  2. Description of the Related Art Conventionally, there are services that publish a website or the like on a communication network such as the Internet and present information desired by a user of the terminal to a terminal that has accessed the website. For example, in a service form such as online shopping, a database in which product information is registered in advance using a predetermined keyword (for example, a search query) designated by the user from the terminal is searched, and the extracted product is used by the user as a purpose. If it is an existing product, the product can be provided to the user by performing the purchase procedure.

  Conventionally, when a search keyword input by the user is input and the “search” button is clicked, not only the search result page is jumped but also other search keywords related to the search keyword input by the user are displayed. There is a mechanism to do this (see, for example, Patent Document 1).

  In Patent Literature 1, a search keyword set input by a user is received, and a plurality of predetermined words corresponding to intentions not clearly defined are compared with the search keyword set, whereby the search keyword set is clearly defined by the user. It is determined whether or not the intention specified in FIG. 4 is indicated, and when it is determined that the search keyword set clearly indicates the intention specified, a process for acquiring further recommended search keywords is executed.

Special table 2014-501422 gazette

  In the above-described method, it is essential to accurately estimate the ambiguity of the intention in order to appropriately switch the processing for displaying the search keyword. Simply, even if you can quantify the ambiguity based on the click distribution for each query, there is actually a difference in the degree of relevance between the clicks, and in order to make a more realistic estimation, Further use of information is required.

  The present invention has been made in view of the above problems, and an object thereof is to quantify the ambiguity of the user's intention and present an appropriate search result to the user.

  The ambiguity evaluation apparatus according to an embodiment of the present invention associates a keyword used at the time of search with a category corresponding to a search result searched from search target data using the keyword, and stores it as search history information. Storage means, and ambiguity evaluation means for calculating similarity between the categories for each keyword from search history information stored in the storage means, and evaluating ambiguity of the keyword from the calculated similarity. .

  According to an embodiment of the present invention, it is possible to quantify the ambiguity of a user's intention and present an appropriate search result to the user.

It is a figure which shows an example of the whole structure of an information presentation system. It is a figure which shows an example of the hardware constitutions of an ambiguity evaluation apparatus. It is a flowchart which shows an example of an information presentation process. It is a figure for demonstrating the ambiguity of a query. It is a figure for demonstrating the relationship of the similarity and ambiguity with respect to a category. It is a figure which shows an example of the calculation result of similarity. It is a figure which shows an example of the search content using fuzziness data.

  Hereinafter, details of each embodiment of the present invention will be described with reference to the accompanying drawings.

<Overall configuration of information presentation system>
FIG. 1 is a diagram illustrating an example of the overall configuration of the information presentation system. In the example of FIG. 1, the information presentation system 10 includes an ambiguity evaluation device 11 that is an example of an information presentation device, and one or a plurality of terminals 12-1 to 12-n (hereinafter, “terminal 12” as necessary). Generically). The ambiguity evaluation device 11 and the terminal 12 are connected to a communication network 13 represented by a LAN (Local Access Area) or the Internet in a state where data can be transmitted and received.

  The ambiguity evaluation apparatus 11 publishes Web sites such as shopping sites, news sites, Q & A sites, electronic commerce sites, and search sites. The ambiguity evaluation apparatus 11 receives input of a predetermined keyword (for example, a search query) from the terminal 12 accessed via the communication network 13, and stores search target data (for example, product data) stored in advance. , The corresponding product information (title, category, etc.), search results such as words, sentences, etc. are extracted and provided to the terminal 12 that has entered the keywords. In addition, the ambiguity evaluation apparatus 11 associates the search query from each terminal 12 with the category for the search result searched from the product data using the search query, and stores it, for example, as a search log (search history information). .

  Here, the category is an individual classification when things are classified based on a certain standard, and is classified by, for example, classification, field, classification, type, department, genre, and the like. For example, when the search query is “table”, the category is “low table”, “center table”, “side table”, etc., but is not limited thereto. In the present embodiment, the ambiguity evaluation device 11 may acquire the above-described product data and search log information from an external device that provides another Web site.

  Further, the ambiguity evaluation device 11 calculates the similarity between categories using the above-described product data and search log, and evaluates the ambiguity related to the keyword (query) based on the calculated similarity. For example, the ambiguity evaluation apparatus 11 generates a vector expression for each category from the product data, extracts a category distribution for each query from the click result of the search log, and obtains the vector expression and the category distribution described above. The ambiguity for each query is evaluated based on the similarity between categories. In addition, the ambiguity evaluation apparatus 11 may calculate the similarity based on the category tree when the data has a tree structure between categories with high reliability (category tree) in advance. In this case, for example, with respect to a keyword (for example, a table) used at the time of search, the tree structure is referred to, and a similarity is calculated from a relationship such as a category (for example, a row table) of a lower concept of a category corresponding to the keyword. can do.

  Further, when the ambiguity evaluation apparatus 11 receives a search query from the terminal 12, the ambiguity evaluation apparatus 11 refers to the ambiguity evaluation result (ambiguity data) and determines other information (for example, based on the ambiguity for the search query). The search query is supplemented by another search query) and the search is performed with the supplemented contents. Further, the ambiguity evaluation apparatus 11 presents the search result to the terminal 12.

  The ambiguity evaluation device 11 may be, for example, a general-purpose PC (Personal Computer) or a server, or may be a cloud server configured by cloud computing having at least one information processing device, but is not limited thereto. .

  The terminal 12 is a device for each user to receive a service from the ambiguity evaluation device 11 or the like. The user accesses the ambiguity evaluation apparatus 11 from the terminal 12 via the communication network 13 and searches for desired information (product information) using a search screen or the like published on the website. When performing a search, by performing user registration or the like, history information for each user can be acquired, and an appropriate search result can be presented for each user. The terminal 12 may be a PC or a communication terminal such as a tablet terminal, a smartphone, or a mobile phone, but is not limited thereto.

  As described above, the communication network 13 is a LAN, the Internet, or the like. The communication network 13 may be wired, wireless, or a combination thereof.

<Functional Configuration of Ambiguity Evaluation Device 11>
Next, an example of a functional configuration of the above-described ambiguity evaluation apparatus 11 will be specifically described. As shown in FIG. 1, the ambiguity evaluation apparatus 11 includes an input means 21, an output means 22, a storage means 23, an ambiguity evaluation means 24, a search means 25, an information presentation means 26, and a communication means 27. And control means 28.

  The input unit 21 accepts inputs such as ambiguity evaluation processing in this embodiment, start / end of information presentation processing, and various settings. For example, if the ambiguity evaluation apparatus 11 is a general-purpose computer such as a PC, the input unit 21 is a pointing device such as a keyboard or a mouse. Further, the input means 21 may be a touch panel or the like. Further, the input means 21 may be a voice input device such as a microphone that can input the above-described voice, for example.

  The output unit 22 outputs the content input by the input unit 21 and the content executed based on the input content. The output means 22 is, for example, a display or a speaker. The output means 22 may be a touch panel integrated with the input means 21. The output unit 22 may be a printing device such as a printer that prints a processing result or the like on a printing medium such as paper.

  The storage unit 23 stores various information necessary for the present embodiment. Specifically, the storage unit 23 stores the product data 31 for extracting product information corresponding to the input search query, the search query, and the category corresponding to the search result by the search query in association with each other. 32, ambiguity data 33 storing the ambiguity for each search query, information presentation data 34 storing supplementary information and presentation information corresponding to the ambiguity, and the like. The product data 31 is an example of search target data, and is not limited to product data.

  Further, the information stored in the storage unit 23 is not limited to these, and for example, a category associated with a keyword used at the time of searching is stored as data (category tree) having a tree structure between categories set in advance. May be. Further, the storage unit 23 may store, for example, address information for communicating with the terminal 12, user information for identifying the user, terminal information for identifying the terminal 12, abnormal information when an error occurs, and the like. Good.

  The ambiguity evaluation means 24 evaluates the ambiguity for each search query based on the similarity between categories obtained as a result of the search query. For example, the ambiguity evaluation means 24 replaces each category with a vector in a linguistic space when calculating the similarity of the category.

  For example, the ambiguity evaluation means 24 generates a vector representation for each category from the product data 31. For the generation of the vector representation, for example, a technique such as a latent semantic index (LSI) can be used, but the present invention is not limited to this. Further, the ambiguity evaluation means 24 extracts the category distribution (click rate) for each query from the number for each category associated with each search query included in the search log 32. Further, the ambiguity evaluation means 24 calculates the similarity by combining the vector expression and the category distribution, and evaluates the ambiguity for each query based on the calculated similarity.

  It should be noted that the search log 32 may be a collection of logs of each terminal 12, may be for each terminal 12 (for each user), and is classified according to predetermined conditions (for example, period, time zone, region) and the like. It may be. The search log 32 may store a category group clicked (browsing instruction) by the user among the categories corresponding to the search result. In this case, the ambiguity evaluation means 24 extracts the category distribution (click ratio) for each query from the number of clicks (viewing instruction number) for each category included in the search log 32, and combines the vector expression and the category distribution for similarity. Calculate the degree. Thus, by reducing the amount of information in the search log 32, it is possible to speed up the processing. In addition, what the user has instructed to browse is likely to be the information desired by the user, so by calculating the similarity using only the useful information, appropriately evaluate the ambiguity for each query, The evaluation result can be used for the next search.

  Note that the ambiguity evaluation unit 24 compares the metadata such as words, genres, attribute information, and detailed explanations associated with each category in the above-described similarity calculation, and the number and types of matching data. The similarity may be calculated by

  For example, when a preset category tree is stored in the storage unit 23 or the like, the ambiguity evaluation unit 24 may calculate the similarity based on the category tree. The category tree is a tree structure in which relationships between categories are formed, such as “table> center table> glass table”, “table> center table> wooden table”, and the like. For example, when there is a tree structure such as “table> row table”, “table> center table”, “table> side table”, “row table”, “center table”, “ The similarity between “side tables” is calculated as high.

  The category tree preferably has a highly reliable (sufficiently reliable) tree structure. The tree structure is usually edited manually. For example, the category “digital camera” is classified into another category by a person such as “PC-related equipment> digital camera”, “camera> digital camera”, and the like. In such a case, the reliability is low. The similarity can be calculated with high accuracy by using a highly reliable category tree that is mechanically formed based on a predetermined condition, instead of the category tree having low reliability.

  The ambiguity evaluation unit 24 stores the ambiguity evaluation result for each query obtained based on the above-described similarity between categories as the ambiguity data 33 in the storage unit 23.

  When the search means 25 receives an input of a search query from the terminal 12, the search means 25 refers to the ambiguity data 33 corresponding to the search query, and if the search query is not ambiguous, the search data 25 uses the search query as it is. A search to 31 is performed. In addition, when the search query is ambiguous, the search unit 25 performs a search query replenishment process and performs a search using the supplemented contents.

  Here, the supplement process is, for example, adding another search query set in advance corresponding to the input search query to expand the search range, or preregistered user preference information (for example, sports enthusiasts) , Gardening enthusiasts, fashion enthusiasts) or the like, or a process of narrowing the search range by changing the input query, but is not limited to this. The other search queries and user preference information described above can be acquired with reference to the information presentation data 34, for example. The search unit 25 stores the search query and the category corresponding to the search result in the search log 32 in association with each other.

  The information presenting means 26 presents the result searched by the searching means 25 to the terminal 12. Note that the information presenting means 26 may arrange and display the search results in the order corresponding to the user preferences when the user preference information and the like are registered in advance when presenting the search results. Further, the information presentation unit 26 may store the information in the search log 32 when there is a browsing instruction such as a click on the category of the search result from the presented terminal 12.

  The communication unit 27 transmits / receives data to / from each terminal 12 via the communication network 13. Note that the communication unit 27 may perform user authentication processing when accessing from the terminal 12. With the user authentication process, for example, it is determined whether authentication information such as a user ID or password input from the terminal 12 matches the user ID or password included in the user information stored in the storage unit 23 in advance. If they match, access from the terminal 12 is allowed. If they do not match, the user is notified of this, and unless the user newly registers, this processing is not allowed to be used by the user. Is not to be done.

  The control means 28 controls the entire components of the ambiguity evaluation apparatus 11. For example, the control unit 28 controls the ambiguity evaluation for each search query by the ambiguity evaluation unit 24, the search by the search unit 25, the information presentation by the information presentation unit 26, etc., but is not limited thereto. . For example, the control unit 28 may perform control such as start and end of various processes in the present embodiment, control when an error occurs, and the like.

<Hardware Configuration of Ambiguity Evaluation Device 11>
Here, in the present embodiment, an execution program (ambiguity evaluation program) that causes a computer to execute each function of the above-described ambiguity evaluation apparatus 11 is generated and installed in, for example, a general-purpose PC or server. An information presentation process or the like in the form can be realized.

  FIG. 2 is a diagram illustrating an example of a hardware configuration of the ambiguity evaluation apparatus. The computer body of the ambiguity evaluation apparatus 11 shown in FIG. 2 includes an input device 41, an output device 42, a drive device 43, an auxiliary storage device 44, a memory device 45, and a CPU (Central Processing Unit) that performs various controls. ) 46 and a network connection device 47, which are connected to each other via a system bus B.

  The input device 41 has a pointing device such as a keyboard and a mouse operated by a user, and has a voice input device such as a microphone, and inputs various operation signals such as execution of a program from the user.

  The output device 42 has a display that displays various windows and data necessary for operating the computer main body that performs each process in the present embodiment, and displays the execution progress, results, and the like by a control program that the CPU 46 has.

  Here, the execution program installed in the computer main body in the present embodiment is provided by, for example, a portable recording medium 48 such as a USB (Universal Serial Bus) memory or a CD-ROM. The recording medium 48 can be set in the drive device 43, and an execution program included in the recording medium 48 is installed in the auxiliary storage device 44 from the recording medium 48 via the drive device 43.

  The auxiliary storage device 44 is a storage unit such as a hard disk, and stores the execution program in the present embodiment, a control program provided in the computer, and the like, and can perform input / output as necessary.

  The memory device 45 stores an execution program read from the auxiliary storage device 44 by the CPU 46. The memory device 45 is a ROM (Read Only Memory), a RAM (Random Access Memory), or the like. Note that the auxiliary storage device 44 and the memory device 45 described above may be configured integrally as a single storage device.

  The CPU 46 controls processing of the entire computer, such as various operations and input / output of data with each hardware component, based on a control program such as an OS (Operating System) and an execution program stored in the memory device 45. Thus, the information presentation process in the present embodiment is realized. Various information necessary during the execution of the program may be acquired from the auxiliary storage device 44 and the execution result or the like may be stored.

  The network connection device 47 acquires an execution program from another device connected to the communication network by connecting to a communication network such as the Internet or a LAN. Further, the network connection device 47 can provide the execution result obtained by executing the program or the execution program itself in the present embodiment to another device or the like.

  The recording medium 48 is a computer-readable recording medium that stores an execution program and the like as described above. The recording medium 48 may be a semiconductor memory such as a flash memory, for example. The recording medium 48 may be a portable recording medium such as a USB memory, but is not limited thereto.

  In this embodiment, by installing an execution program (for example, an ambiguity evaluation program or the like) in the hardware configuration of the computer main body described above, the ambiguity evaluation processing in the present embodiment is performed in cooperation with hardware resources and software. Etc. can be realized. Further, the ambiguity evaluation program corresponding to the ambiguity evaluation process described above may be resident on the apparatus, for example, or may be activated by an activation instruction. The hardware configuration described above may be applied to the terminal 12, for example.

<Information presentation process in this embodiment>
Next, information presentation processing including ambiguity evaluation processing in the present embodiment will be described using a flowchart. FIG. 3 is a flowchart illustrating an example of the information presentation process. In the following description, the ambiguity evaluation apparatus 11 has already been searched by each terminal 12 and the like, and a search log exists.

  In the example of FIG. 3, the ambiguity evaluation means 24 generates a vector representation for each category from the product data 31 as an example of search target data stored in advance (S01). Next, the ambiguity evaluation means 24 extracts the category distribution for each query from the click result of the search log 32 (S02). In the process of S02, the category distribution may be extracted using the category of the search result instead of the click result for the category of the search result included in the search log 32.

  Next, the ambiguity evaluation unit 24 calculates the similarity between categories for each query by combining the vector expression and the category distribution (S03), evaluates the ambiguity for each query from the calculated similarity, and determines the ambiguity. The data 33 is stored in the storage means 23 (S04). Note that the processing of S01 to S04 only needs to be performed prior to the processing after S05, and does not need to be performed continuously.

  Next, the search means 25 receives the input of the search query from the terminal 12 (S05), refers to the ambiguity data for the received search query, and acquires the corresponding ambiguity (S06). Next, the search means 25 determines whether or not the acquired ambiguity is greater than or equal to a preset threshold value (S07). If it is greater than or equal to the threshold value (YES in S07), other information (for example, The search is performed by supplementing additional keywords (S08). If the threshold value is not greater than or equal to the threshold value in the process of S07 (NO in S07), the search unit 25 performs a search by a search query without performing supplementation (S09).

  After the processing of S08 or S09, the information presenting means 26 presents information of the search result (S10). Further, the information presentation unit 26 acquires information such as a click result from the terminal 12 for the presented search result, and stores it in the search log 32 together with the search query, the search result, and the like (S11).

  Here, the search unit 25 determines whether or not an input of another search query has been accepted (S12), and when an input of another search query has been accepted (YES in S12), the process returns to the process of S06. Further, when the input of another search query is not accepted (NO in S12), for example, after a predetermined time has elapsed, or when there is processing such as closing the website screen from the terminal 12, the control means 28 Then, the information presentation process in this embodiment is terminated.

<Query ambiguity>
Here, the ambiguity of the query in the present embodiment will be described with reference to the drawings. FIG. 4 is a diagram for explaining the ambiguity of the query.

  For example, in the example shown in FIG. 4A, when the category clicked by the user is “low table”, “center table”, “side table”, etc., among the search results for the search query “table”, these are: Since it corresponds to the genre (attribute) of the same product “desk”, the degree of similarity is high. Therefore, if the similarity of each category is high (for example, it is equal to or higher than a preset threshold), it is determined that the ambiguity of the query is low, and the search is performed without performing the replenishment process for the query of “table”. Is done.

  In the example shown in FIG. 4B, it is assumed that the category clicked by the user among the search results for the search query “balloon” is “balloon bowl”, “balloon skirt”, and “balloon telegram”. In such a case, although each has the letter “balloon”, the genre of the product is not the same. Therefore, in such a case, the similarity between the categories is low, and the query “balloon” performs the above-described replenishment process as a keyword with high ambiguity. For example, when there is preference information such as “gardening preference” as user information, the search query is searched with a compound word such as “balloon × gardening”. Thereby, an appropriate search result can be presented to the user.

<Ambiguity assessment method>
Next, an example of the ambiguity evaluation method in the present embodiment will be specifically described. The ambiguity evaluation means 24, for example, information on the product category and product title obtained from the product data 31, the search query included in the search log 32, and the click of each user for each category of the search result by the search query The ambiguity of the search query is determined by counting the number (number of browsing instructions).

  For example, since the association between the category and the product is usually performed based on a person's judgment or the like, it has imperfections. In addition, the categories are not necessarily independent, and there are categories with high similarity to other categories. Therefore, in this embodiment, each category is mapped to a low-dimensional vector space using an LSI or the like. LSI is usually applied to a relationship matrix between terms (words) and sentences (documents), but in this embodiment, it is applied to a relationship matrix between queries and categories.

  For example, in the present embodiment, a query / category association matrix is configured using scores based on TF (Term Frequency) -IDF (Inverse Document Frequency) as shown in the following formula (1).

ここで、上述した式（１）において、ｔは単語（ターム）を示し、ｃはカテゴリを示し、ｄはドキュメントを示し、商品データにおけるタイトルに相当する。また、ｄｏｃｓ（ｃ）はカテゴリｃに含まれているドキュメント（商品）の集合を示す。また、ｔｅｒｍｓ（ｄ）はドキュメントの中に含まれている単語の集合を示す。また、ｎ_ｔ，ｄは、単語ｔが文章（商品）ｄの中で何回使用されているかを示す。また、Ｄはドキュメント全体の集合を示し、｜Ｄ｜は、ドキュメント全体の数（全商品数）を示す。なお、（１）式を用いた計算において、カテゴリ数は少ないため、通常のターム（単語）と文章（ドキュメント）とを用いたスコア計算よりも迅速に計算することができる。

Here, in the above formula (1), t indicates a word (term), c indicates a category, d indicates a document, and corresponds to a title in product data. Docs (c) indicates a set of documents (products) included in the category c. Further, terms (d) indicates a set of words included in the document. N _{t, d} indicates how many times the word t is used in the sentence (product) d. D indicates the set of the entire document, and | D | indicates the total number of documents (the total number of products). In the calculation using equation (1), since the number of categories is small, it can be calculated more quickly than the score calculation using normal terms (words) and sentences (documents).

  In the present embodiment, the ambiguity with respect to the query is quantified by aggregating the category vector c based on the “direction” of the vector. FIG. 5 is a diagram for explaining the relationship between similarity and ambiguity with respect to a category. For example, FIGS. 5A and 5B show linguistic space vectors, and c1 to c3 indicate categories extracted by the search query and clicked by the user.

  When this category is dropped into the spaces shown in FIGS. 5A and 5B, the direction of each category is discrete in the example of FIG. Become. Therefore, in the case of FIG. 5A, it is evaluated that the similarity between categories is low and the ambiguity of the query is high. In the example of FIG. 5B, the directions of the categories are close to each other, and the aggregation result of the vectors is a large vector amount. Therefore, in the case of FIG. 5B, it is evaluated that the similarity between categories is high and the ambiguity of the query is low.

  In this embodiment, in order to evaluate the uniformity of the vector direction, first, the center of the direction (center of gravity) for each vector is acquired. In the present embodiment, for example, the center of the category direction can be calculated by the following equation (2).

ここで、上述した式（２）において、Ｃ_ｌｅａｆは、階層構造を有するカテゴリｃの葉（末端）の部分である。なお、本実施形態では、末端カテゴリＣ_ｌｅａｆに限定した例を用いて説明するが、これに限定されるものではなく、例えば全カテゴリＣを用いてもよい。また、ｐ（ｃ｜ｑ）は、クエリｑが与えられたときに、カテゴリｃが出てくる確率を示しており、ｃ∈Ｃ_ｌｅａｆの関係を有する。また、末端カテゴリＣ_ｌｅａｆに限定しない場合には、ｃ∈Ｃの関係を有する。

Here, in the above-described equation (2), C _leaf is a leaf (terminal) portion of category c having a hierarchical structure. In the present embodiment, an example in which the terminal category C _leaf is limited will be described. However, the present invention is not limited thereto, and for example, all categories C may be used. Further, p (c | q) indicates the probability that the category c appears when the query q is given, and has a relationship of c∈C _leaf . Moreover, when it does not limit to the terminal category C _leaf , it has the relationship of c∈C.

  Next, the similarity of each category can be calculated by calculating the distance of each category vector based on the acquired center using the cosine similarity or the like. For example, the following formula (3) can be used for calculating the similarity.

ここで、上述した式（３）において、クエリｑの類似度ｓｉｍ（ｑ）は、０から１の範囲で与えられ、類似性が高いほど、値が１に近くなり、類似性が低いほど値が０に近くなる。

Here, in the above-described equation (3), the similarity sim (q) of the query q is given in the range of 0 to 1, and the higher the similarity is, the closer the value is to 1, and the lower the similarity is the value. Is close to zero.

<Example of similarity calculation result>
Here, an example of the similarity calculation result will be described with reference to the drawings. FIG. 6 is a diagram illustrating an example of a similarity calculation result. In the example of FIG. 6, the similarity is calculated from a search query (keyword) used on a website such as an actual commercial transaction site, a search result (number of clicks), or the like.

  In the example of FIG. 6, with respect to two examples of “table” and “balloon” as queries, each category acquired from the search log 32 and a click (viewing instruction) ratio for each category ( probability) and the calculation result (sim) by the above formula (3) for each query are shown in the example of Fig. 6. In the case of the query "table", the click rate for all 265 categories. The maximum value of (probability) is 0.094, and the other five top categories are shown. Almost all categories including the top five categories are included in the same genre and attribute as the table (Table), and the similarity (sim) according to the above-described equation (3) is also a high value of 0.883. It was.

  On the other hand, in the example of FIG. 6, in the case of the query “balloon”, the maximum click ratio is 0.094 for all 152 categories, and the other five categories are shown. The top five categories include those with different genres and attributes such as gardening, toys, fashion, etc., and the similarity (sim) according to the above equation (3) is also 0. It was a low figure of 376.

  In the present embodiment, ambiguity is quantified using the above-described similarity and stored as ambiguity data 33. As a result, for the subsequent search request, the ambiguity for the input search query is determined from the ambiguity data 33, and when the similarity is low (the ambiguity is high), the search keyword is expanded and the user information is used. To supplement the query, perform the search process, and present the results.

<Search contents using ambiguity data>
Next, search contents using ambiguity data in the present embodiment will be described with reference to the drawings. FIG. 7 is a diagram showing an example of search contents using ambiguous data. In the example of FIG. 7, ambiguity data is generated and search contents using the generated ambiguity data are shown.

  In the example of FIG. 7, as described above, the ambiguity with respect to the query is evaluated using the product data 31 and the search log 32. In the present embodiment, as shown in (1) of FIG. 7, a vector representation for each category is generated from the category and title (product name) included in the product data 31. For example, LSI or the like can be used for the vector expression, and the direction and amount of each category vector can be calculated using the above-described equation (1).

  In this embodiment, as shown in (2) of FIG. 7, the distribution (click ratio) of each category clicked (viewed) on a query (for example, a table) is extracted using the search log 32.

  Next, the similarity between categories is calculated using each category vector based on the distribution of each extracted category, and the ambiguity is evaluated from the calculated result. In the example of FIG. 7, using the category vectors corresponding to “low table”, “center table”, and “side table”, the similarity is calculated by the above formulas (2) and (3), and the calculation result is The ambiguity evaluation result is stored in the ambiguity data 33. In the example of (3) in FIG. 7, the similarity “0.88” obtained by Expression (3) is stored in association with the query (table) as the evaluation result of the ambiguity. Note that the data stored in the ambiguity data 33 is not limited to the above-described calculation result of the similarity, but for example, a value calculated using the similarity as “ambiguity = 1−similarity” ( For example, ambiguity = 1-0.88 = 0.12) may be stored in the ambiguity data 33.

  By using this ambiguity data 33, for example, as shown in FIG. 7 (4), when a search query of “table” is input in the search site provided by the ambiguity evaluation apparatus 11, the ambiguity data 33 Since the ambiguity with respect to the table is low, the search is performed with the query as it is. In the case of another search query “balloon”, since the similarity is as low as 0.376 (high ambiguity) in the example of FIG. 6, another search keyword is added or user information is used. The search is performed by supplementing the keyword corresponding to the preference information, and the search result is presented.

<Summary>
As described above, according to the present embodiment, the ambiguity of the user's intention can be quantified. Thereby, this embodiment can present an appropriate search result to a user. In the above-described embodiment, the query for which the ambiguity is evaluated is one word. However, even in the case of a compound word composed of a plurality of words, the compound word is used as one search query between the above categories. Similarity can be calculated, and ambiguity can be evaluated based on the calculated similarity. As for the above-described similarity calculation between categories, the similarity may be calculated by replacing it with a vector in a linguistic space, or the similarity is calculated using a tree structure of a preset category. These may be combined.

  The preferred embodiments of the present invention have been described in detail above, but the present invention is not limited to such specific embodiments, and various modifications can be made within the scope of the gist of the present invention described in the claims. Can be changed.

  In addition, the order of the flowcharts described in the above embodiments may be changed as long as there is no contradiction. Moreover, all or some of the above-described embodiments can be combined. In addition, all or part of the above-described embodiments can be implemented by a program. This program can be stored in a storage medium.

  Note that the present invention is not limited to the configurations shown here, such as combinations with other elements, etc., in the configurations described in the above embodiments. These points can be changed without departing from the spirit of the present invention, and can be appropriately determined according to the application form.

DESCRIPTION OF SYMBOLS 10 Information presentation system 11 Ambiguity evaluation apparatus 12 Terminal 13 Communication network 21 Input means 22 Output means 23 Storage means 24 Ambiguity evaluation means 25 Search means 26 Information presentation means 27 Communication means 28 Control means 31 Product data 32 Search log 33 Ambiguity Data 34 Information presentation data 41 Input device 42 Output device 43 Drive device 44 Auxiliary storage device 45 Memory device 46 CPU
47 Network connection device 48 Recording medium

Claims (10)

Storage means for associating and storing as a search history information a keyword used at the time of searching and a category corresponding to a search result searched from search target data using the keyword;
From the search history information stored in the storage means, the similarity between the categories for each keyword is calculated, and the ambiguity evaluation means for evaluating the ambiguity of the keyword from the calculated similarity,
The ambiguity evaluation apparatus characterized by having. The ambiguity evaluation means includes
Generating a spatial vector representation for each of the categories included in the search target data;
Extracting the category distribution for each keyword based on the number for each category obtained from the search history information,
The ambiguity evaluation apparatus according to claim 1, wherein the similarity is calculated based on the vector expression and the category distribution. The ambiguity evaluation means includes
The ambiguity evaluation apparatus according to claim 1 or 2, wherein ambiguity data corresponding to the similarity is stored in the storage unit for each keyword.   A search means is provided for acquiring a ambiguity for the keyword from the ambiguity data with respect to a keyword input for a new search, and performing a search corresponding to the acquired ambiguity. Item 5. The ambiguity evaluation apparatus according to Item 3. The search means includes
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 5. The ambiguity evaluation apparatus according to claim 4, wherein if it is not the above, a search is performed using the keyword. The storage means
6. The category corresponding to a search result searched from the search target data, the category designated by the user for browsing is stored as search history information in association with the keyword. The ambiguity evaluation apparatus described in 1. The ambiguity evaluation means includes
The ambiguity evaluation apparatus according to any one of claims 1 to 6, wherein the similarity is calculated based on data of a tree structure between the categories set in advance. Storage means for storing a category associated with a keyword used at the time of search as data of a tree structure between preset categories;
Ambiguity evaluation means for calculating the similarity between the categories for each keyword based on the data of the tree structure between the categories stored in the storage means, and evaluating the ambiguity of the keyword from the calculated similarity; ,
The ambiguity evaluation apparatus characterized by having. A storage step of associating the keyword used at the time of searching with the category corresponding to the search result searched from the search target data using the keyword, and storing it in the storage means as search history information;
From the search history information stored in the storage means, the similarity between the categories for each keyword is calculated, and the ambiguity evaluation step of evaluating the ambiguity of the keyword from the calculated similarity,
Ambiguity Evaluation Method that Causes a Computer to Run Computer
Storage means for associating and storing as a search history information a keyword used at the time of searching and a category corresponding to a search result searched from search target data using the keyword;
From the search history information stored in the storage means, the similarity between the categories for each keyword is calculated, and the ambiguity evaluation means for evaluating the ambiguity of the keyword from the calculated similarity,
Ambiguity evaluation program to function as

Images