JP7067884B2 - Classification device, classification method and classification program - Google Patents

Description

The present invention relates to a classification device, a classification method and a classification program.

In recent years, with the development of communication networks, various services have been provided via networks. In connection with this, various search techniques for searching for services existing on the network and searching for desired products and articles within the services have been proposed.

For example, a search technique using a concept keyword extended data set generated by classifying a modifier and a concept keyword from a search keyword transmitted from a user terminal and extracting each of the modifier and the concept keyword has been proposed. There is.

Japanese Unexamined Patent Publication No. 2013-73626

However, with the above-mentioned conventional technique, it is difficult to derive mutual relationships between various queries. Specifically, in the above-mentioned conventional technique, keyword expansion processing is performed based on words, features, and the like common to a plurality of queries (keywords) input by the user. That is, in the prior art, it is not possible to extract common features from various queries whose fields and categories are different from each other, and as a result, information processing such as keyword expansion and analysis of the relationship between queries is performed. May not be possible.

The present application has been made in view of the above, and an object thereof is to provide a classification device, a classification method, and a classification program capable of deriving mutual relationships between various queries.

The classification device according to the present application is based on an extraction unit that extracts a plurality of second queries related to the first query based on the degree of relevance between arbitrary queries, and a plurality of second queries extracted by the extraction unit. It is characterized by including a generation unit that generates feature information that characterizes the first query, and a classification unit that classifies keywords corresponding to the first query based on the feature information generated by the generation unit. And.

According to one aspect of the embodiment, there is an effect that the mutual relationship between various queries can be derived.

FIG. 1 is a diagram showing an example of a classification process according to an embodiment. FIG. 2 is a diagram showing a configuration example of a classification system according to an embodiment. FIG. 3 is a diagram showing a configuration example of the classification device according to the embodiment. FIG. 4 is a diagram showing an example of a relevance information storage unit according to an embodiment. FIG. 5 is a diagram showing an example of a feature information storage unit according to an embodiment. FIG. 6 is a diagram showing an example of a classification information storage unit according to an embodiment. FIG. 7 is a flowchart (1) showing a processing procedure by the classification device according to the embodiment. FIG. 8 is a flowchart (2) showing a processing procedure by the classification device according to the embodiment. FIG. 9 is a flowchart showing a processing procedure by the classification device according to the modified example. FIG. 10 is a hardware configuration diagram showing an example of a computer that realizes the function of the classification device.

Hereinafter, the classification device, the classification method, and the embodiment for implementing the classification program (hereinafter referred to as “the embodiment”) according to the present application will be described in detail with reference to the drawings. It should be noted that this embodiment does not limit the classification device, classification method and classification program according to the present application. Further, in each of the following embodiments, the same parts are designated by the same reference numerals, and duplicate explanations are omitted.

[1. An example of classification processing]
First, an example of the classification process according to the embodiment will be described with reference to FIG. FIG. 1 is a diagram showing an example of a classification process according to an embodiment. In FIG. 1, as an example of the classification process according to the embodiment, a process in which the classification device 100 according to the embodiment classifies a plurality of keywords included in the keyword list L01 transmitted from the user U01 will be described as an example.

The classification device 100 shown in FIG. 1 generates feature information of each query based on the degree of relevance between search queries (hereinafter referred to as “query”) transmitted from the user, and is based on the generated feature information. It is a server device that classifies queries into predetermined classes (clustering). In the embodiment, the classification device 100 also functions as a web server that provides a search site to the user.

In the embodiment, a keyword used for a search on a search site or the like is referred to as a "query". However, there may be cases where the keyword and query do not exactly match. For example, the user may enter a sentence such as "AAA car" as a query ("AAA" is, for example, the name of the car manufacturer). In this case, "AAA car" can be a query, and the words "AAA" and "car" included in the query can be keywords. The classification device 100 appropriately extracts keywords from the query by, for example, performing morphological analysis of the query. Further, in the following description, "query" and "keyword" may be treated as having the same meaning. For example, if the query matches the keyword, or if one keyword is extracted from the result of morphological analysis of the query, the query and the keyword are treated as synonymous.

The user group shown in FIG. 1 is a plurality of users who use a search site or the like, and are a plurality of users who send a query to the classification device 100. Further, the user U01 is an example of a user, and is a user who transmits a keyword list L01 including a plurality of keywords to the classification device 100 and requests classification of keywords included in the keyword list L01. Although not shown in FIG. 1, each user and each user U01 included in the user group uses an information processing terminal for using a search site and transmitting / receiving various information to / from the classification device 100. Hereinafter, it is referred to as "user terminal 10"). Further, in the following description, the user terminal 10 and the user who uses the user terminal 10 may be equated with each other. For example, "user U01 sends a query" actually means "user terminal 10 used by user U01 sends a query".

The query that the user sends when performing a search indicates the user's interests. Therefore, by classifying the queries, for example, it is possible to presume that users who frequently send queries belonging to the same class are users who have the same interests as each other. Such information can be useful information in marketing such as advertisement distribution, for example. From this, if it is possible to find out what kind of queries have a similar relationship with each other and classify the queries, for example, useful information for an advertisement distribution business operator or the like can be obtained. As a method of classifying queries, for example, there is a method of setting a category for a query and classifying queries belonging to the same category into the same class.

However, the types of queries sent to search sites and the like are enormous, and it is not realistic to manually assign categories. Further, even in the method of assigning a category to a query by a program or the like, one query may contain a plurality of meanings, and it is difficult to automatically perform appropriate categorization.

Therefore, the classification device 100 according to the embodiment calculates the degree of relevance between two queries based on the number of times that the same user inputs a different query among all the users to be aggregated. Further, the classification device 100 calculates the feature information of an arbitrary query based on the degree of relevance. Then, the classification device 100 classifies the query based on the feature information. As a result, the classification device 100 can classify the queries based on the quantitative information of the behavior actually searched by the user, so that the related queries can be classified into appropriate classes without categorizing each query. can. Further, according to the classification device 100, it is not necessary to categorize the queries in advance, so that no matter what kind of query is input, it is possible to calculate the degree of relevance between the queries, classify the queries, and the like. can. That is, the classification device 100 can perform classification processing corresponding to various queries. Hereinafter, an example of the classification process performed by the classification device 100 will be described along the flow with reference to FIG. 1. In the following description, of the two queries for which the degree of relevance is calculated, the query to be processed is referred to as a "first query", and the query related to the first query is referred to as a "second query". Therefore, the predetermined keyword may be the "first query" or the "second query".

In the example shown in FIG. 1, each user of the user group inputs a query on a search site or the like (step S11). The classification device 100 acquires a query sent from each user (step S12). Then, the classification device 100 calculates the degree of relevance between the queries (step S13).

The classification device 100 is based on at least the number of users who have entered one of two different queries among the arbitrary queries and the number of users who have entered both of the two queries. Calculate the degree of relevance of. As an example, the classification device 100 relates the first query (“query A” in the formula (1)) and the second query (“query B” in the formula (1)) based on the following formula (1). Calculate the degree.

In the above equation (1), "Score (A, B)" indicates a numerical value of the degree of association between query A and query B. "Auser" indicates the number of users who searched (input) the query A. “Buser” indicates the number of users who searched for query B. "ALLuser" indicates the number of users (users who used the search) who sent a query in a predetermined aggregation period. Then, "User ∧ Buser" indicates the number of users who searched for both Query A and Query B.

The classification device 100 calculates the relevance of all the queries searched from the user group by using the above formula (1). For example, the classification device 100 aggregates the queries B searched by the user who searched the query A, which is the first query. Then, the classification device 100 calculates the degree of relevance for all the second queries for the first query by using the above equation (1). The classification device 100 stores the calculated relevance of the first query in the storage unit. The classification device 100 may extract only the pair of the first query and the second query having a degree of relevance exceeding a predetermined threshold value and store them in the storage unit.

In the above formula (1), the classification device 100 can adjust how long the period of relevance between queries is calculated by changing the aggregation period. For example, if the aggregation period is set within a range of several years, the classification device 100 will "first query and second query" when the same user searches for the first query and the second query during the few years. Since it is counted as "users who searched both queries", it is possible to calculate the degree of relevance indicating the change of interests of users over a relatively long range. As a result, the classification device 100 can calculate the degree of relevance that reflects the transition of the user's life stage (for example, the query searched by the user has changed from "pregnancy" to "childbirth"). On the other hand, if the aggregation period is set within a range of several days, the classification device 100 will not search for both the first query and the second query within the few days, "the first query and Do not count the number of users who searched both in the second query. Therefore, the classification device 100 can easily extract pairs and the like of related queries without requiring a relatively long period of time. The aggregation period may be appropriately set by, for example, the administrator of the classification device 100 or the like.

In the example of FIG. 1, the classification device 100 stores the calculated relevance degree in the database DB 01. As shown in FIG. 1, the database DB 01 stores the degree of relevance when the first query is "AAA" and the second query is "BBB", "CCC", and "DDD", respectively. Specifically, as an example of the information stored in the database DB01, the relevance of the pair in which the first query is "AAA" and the second query is "BBB" is "5.93". Shows.

Subsequently, the classification device 100 performs a process of generating feature information of each query (keyword). First, the classification device 100 extracts a second query related to the first query based on the calculated relevance degree (step S14). Specifically, the classification device 100 extracts all the second queries whose relevance to the first query exceeds a predetermined threshold. For example, the classification device 100 uses a predetermined search engine that performs a process of extracting a second query whose relevance to the first query exceeds a predetermined threshold when the first query is input. Extract.

Then, the classification device 100 generates feature information indicating the features of the first query based on the extracted second query (step S15). For example, the classification device 100 generates the feature information of the first query as a vector by expressing the extracted second query as a word vector having each dimension. In this case, the classification device 100 may perform morphological analysis of each of the extracted second queries and generate a word vector indicating the first query based on the analyzed keywords.

As mentioned above, a query may contain multiple keywords. Suppose that the query "BBB used car" exists as the second query related to the first query. In this case, the classification device 100 performs morphological analysis of "used car of BBB" and extracts each keyword of "BBB", "used", and "car". The classification device 100 acquires each keyword that characterizes the first query and the number of occurrences of the keyword by performing the above processing for all the extracted second queries. Then, the classification device 100 generates a vector in which each keyword is a dimension and the number of appearances of each keyword is a dimension number. Temporarily, the above processing is performed for the second query related to the first query "AAA", and "BBB" appears "18" times, "CCC" appears "15" times, and "DDD" as keywords. Appears "9" times, "used" appears "25" times, and "automobile" appears "71" times. In this case, the feature information of the first query "AAA" is as a vector such as (BBB, CCC, DDD, used, automobile, ...) = (18,15,9,25,71, ...). Shown.

The classification device 100 performs the above processing with all the queries searched from the user group as the first query, and generates the feature information of each first query. Then, the classification device 100 stores the first query and the generated feature information in the storage unit (database DB 02 in the example of FIG. 1) in association with each other.

After that, at a predetermined timing, the user U01 who desires to use the classification process by the classification device 100 transmits a keyword list including any plurality of keywords to the classification device 100 (step S16). In the example of FIG. 1, the user U01 transmits a keyword list L01 including a list of names of automobile manufacturers as a keyword to the classification device 100.

The classification device 100 receives the transmitted keyword list L01. Then, the classification device 100 classifies (clusters) the keywords included in the keyword list L01 based on the similarity of the feature information (step S17).

For example, the classification device 100 refers to the database DB 02 and extracts the first query corresponding to each keyword included in the keyword list L01. Then, the classification device 100 classifies the keywords included in the keyword list L01 based on the relevance of each of the extracted feature information of the first query. Specifically, the classification device 100 classifies each keyword included in the keyword list L01 by using a non-hierarchical method such as a k-means method. The method of classification processing is not limited to the above example, and the classification device 100 may use any method as long as it is a method capable of classifying keywords based on feature information. For example, the classification device 100 may calculate the cosine similarity between feature information and classify keywords having cosine similarity exceeding a predetermined threshold into the same class. Further, the classification device 100 may use a hierarchical method such as the shortest distance method, or may use a classification method using learning such as a support vector machine.

In the example shown in FIG. 1, it is assumed that the classification device 100 classifies the keywords "AAA", "BBB", "CCC" and the like included in the keyword list L01 into the class CL01 which is the same class. Further, it is assumed that the classification device 100 classifies the keywords "DDD" and the like included in the keyword list L01 into the class CL02. In the example of FIG. 1, the query "AAA", "BBB", etc. and the keywords "AAA", "BBB", etc. indicate the same word, respectively.

The classification device 100 stores the classification result in a storage unit (database DB 03 in the example of FIG. 1). In this example, the user has some common qualities in "AAA", "BBB" and "CCC" (eg, because they have a common base country, have a capital relationship, and have similar characteristics of the cars they manufacture. It is presumed that they have (such as being compared with each other). In addition, "AAA", "BBB" and "CCC" and "DDD" have some different properties (for example, the countries in which they are based are different, there is no capital relationship, and the automobiles they manufacture do not compete with each other. It is presumed that they have (such as). As described above, according to the classification device 100, clustering is performed using the feature information generated based on the user's search behavior for the plurality of keywords presented by the user U01, so that the user's interests are reflected. Can be classified.

After that, the classification device 100 transmits the result of classifying a plurality of keywords included in the keyword list L01 to the user U01 (step S18). By referring to the classified results, the user U01 can obtain information such as what kind of keywords are related to the user's behavior.

As described above with reference to FIG. 1, the classification device 100 according to the embodiment extracts a plurality of second queries related to the first query based on the degree of relevance between arbitrary queries. Further, the classification device 100 generates feature information that characterizes the first query based on the plurality of extracted second queries. Then, the classification device 100 classifies the keywords corresponding to the first query based on the generated feature information.

That is, according to the classification device 100, the classification process is performed based on the quantitative information of the actual search behavior of the user, regardless of the qualitative information such as the meaning and the property of the query transmitted from the user. In other words, the classification device 100 classifies related keywords without categorizing each query. In this way, the classification device 100 can perform classification for various queries transmitted from the user, and as a result, it is possible to derive mutual relationships between various queries. Hereinafter, the classification system 1 including the classification device 100 and the classification device 100 that perform the above processing will be described in detail.

[2. Classification system configuration]
Next, the configuration of the classification system 1 according to the embodiment will be described with reference to FIG. FIG. 2 is a diagram showing a configuration example of the classification system 1 according to the embodiment. As shown in FIG. 2, the classification system 1 includes a user terminal 10 and a classification device 100. The user terminal 10 and the classification device 100 are connected so as to be communicable by wire or wirelessly via a network N (for example, the Internet) which is a communication network. The number of each device included in the classification system 1 shown in FIG. 2 is not limited to that shown in the figure. For example, the classification system 1 may include a plurality of user terminals 10 and the like.

The user terminal 10 is an information processing device used by the user. For example, the user terminal 10 is an information processing device such as a desktop PC, a notebook PC, a mobile phone such as a smartphone, a tablet terminal, a PDA (Personal Digital Assistant), or a wearable device (Wearable Device). For example, the user terminal 10 accesses the search site according to the operation by the user. Then, the user terminal 10 transmits the query input by the user to the server (in the embodiment, the classification device 100) that provides the search site.

As described above, the classification device 100 generates feature information that characterizes the first query based on the degree of relevance between arbitrary queries, and classifies keywords corresponding to the first query based on the generated feature information. It is a device.

[3. Classification device configuration]
Next, the configuration of the classification device 100 according to the embodiment will be described with reference to FIG. FIG. 3 is a diagram showing a configuration example of the classification device 100 according to the embodiment. As shown in FIG. 3, the classification device 100 includes a communication unit 110, a storage unit 120, and a control unit 130. The classification device 100 includes an input unit (for example, a keyboard, a mouse, etc.) that receives various operations from an administrator or the like who uses the classification device 100, and an output unit (for example, a liquid crystal display, etc.) for outputting various information. You may have.

(About communication unit 110)
The communication unit 110 is realized by, for example, a NIC (Network Interface Card) or the like. The communication unit 110 is connected to the communication network by wire or wirelessly, and transmits / receives information to / from the user terminal 10 via the communication network.

(About the storage unit 120)
The storage unit 120 is realized by, for example, a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory (Flash Memory), or a storage device such as a hard disk or an optical disk. The storage unit 120 according to the embodiment includes a relevance information storage unit 121, a feature information storage unit 122, and a classification information storage unit 123. Hereinafter, each storage unit will be described in order.

(About the relevance information storage unit 121)
The relevance information storage unit 121 stores information regarding the relevance between queries. The relevance information storage unit 121 corresponds to the database DB 01 shown in FIG. Here, FIG. 4 shows an example of the relevance information storage unit 121 according to the embodiment. FIG. 4 is a diagram showing an example of the relevance information storage unit 121 according to the embodiment. As shown in FIG. 4, the relevance information storage unit 121 has items such as "aggregation period", "total number of search users", "first query", "second query", and "relevance".

The "aggregation period" indicates a period during which information regarding the user's search behavior is aggregated. "Total number of search users" indicates information regarding the total number of search users in the aggregation period. In FIG. 4, the information stored in the "total number of search users" is conceptually shown as "A01", but in reality, the item of "total number of search users" is included in the aggregation period. The specific number of all unique users who used the search site is stored. It should be noted that the item may store information such as the number of times a query is sent (in other words, a search is performed) to the search site during the aggregation period, and the number of times each query is searched.

The "first query" indicates the first query. The "second query" is a query searched by the user who searched for the first query, and indicates a query different from the first query. "Relevance degree" indicates the degree of relevance between the first query and the second query. The degree of relevance is calculated for each query pair by, for example, the above equation (1).

That is, in FIG. 4, as an example of the information held by the relevance information storage unit 121, the aggregation period is "July 1, 2016 to June 30, 2017", and the total number of search users is "A01". Of the aggregated data, the degree of relevance of the pair whose first query is "AAA" and whose second query is "BBB" is "5.93".

(About the feature information storage unit 122)
The feature information storage unit 122 stores the feature information of the query. The feature information storage unit 122 corresponds to the database DB 02 shown in FIG. Here, FIG. 5 shows an example of the feature information storage unit 122 according to the embodiment. FIG. 5 is a diagram showing an example of the feature information storage unit 122 according to the embodiment. As shown in FIG. 5, the feature information storage unit 122 has items such as "first query", "extracted second query information", "morphological analysis information", and "feature information".

The "first query" corresponds to the same item shown in FIG. The "extracted second query information" indicates the information of the second query extracted as the second query having a degree of relevance exceeding a predetermined threshold value with respect to the first query. In FIG. 5, the information stored in the "extracted second query information" is conceptually shown as "B01", but in reality, the item of the "extracted second query information" is shown. , Information indicating a plurality of extracted second queries is stored. The classification device 100 may arbitrarily set the threshold value of the degree of relevance when extracting the second query. Further, the classification device 100 may, for example, extract a predetermined number (for example, 10 or 100) of second queries from the first query regardless of the degree of relevance.

"Morphological analysis information" indicates information obtained by morphological analysis of the extracted second query. In FIG. 5, the information stored in the "morphological analysis information" is conceptually shown as "C01", but in reality, the extracted second query is included in the "morphological analysis information" item. The result of morphological analysis is stored. More specifically, in the item of "morphological analysis information", a keyword obtained by morphological analysis of the second query and information indicating the number of occurrences of the keyword are stored. The classification device 100 may store only the keywords satisfying a predetermined condition among the results of the morphological analysis. For example, in the case of Japanese, the classification device 100 may store only nouns as keywords as a result of morphological analysis of the second query.

"Characteristic information" indicates the characteristic information of the first query. In FIG. 5, the information stored in the "feature information" is conceptually shown as "R01", but in reality, the item of the "feature information" is obtained by morphological analysis of the second query. Information composed of the keywords and the number of times the keywords appear is stored. More specifically, in the item of "feature information", a word vector having a keyword as a dimension and the number of occurrences as a dimension is stored.

That is, in FIG. 5, as an example of the information held by the feature information storage unit 122, the second query information extracted for the first query “AAA” is “B01”, and the second query is morphologically analyzed. It is shown that the morphological analysis information is "C01" and the feature information generated from such information is "R01".

(About the classification information storage unit 123)
The classification information storage unit 123 stores the result of the classification process. The classification information storage unit 123 corresponds to the database DB 03 shown in FIG. Here, FIG. 6 shows an example of the classification information storage unit 123 according to the embodiment. FIG. 6 is a diagram showing an example of the classification information storage unit 123 according to the embodiment. As shown in FIG. 6, the classification information storage unit 123 has items such as "keyword list ID", "class ID", and "keyword".

The "keyword list ID" indicates identification information that identifies the keyword list. The "class ID" indicates identification information that identifies the class. In this specification, the identification information such as the keyword list ID is the same as the reference code used in the description. For example, the keyword list identified by the keyword list ID "L01" indicates "keyword list L01". "Keyword" indicates a keyword to be classified.

That is, in FIG. 6, as an example of the information held by the classification information storage unit 123, the keywords included in the keyword list L01 identified by the keyword list ID “L01” are identified by the class IDs “CL01”, “CL02”, and the like. It shows that it was classified into the class to be. Further, in FIG. 6, for example, it is shown that the keyword “AAA”, the keyword “BBB”, and the keyword “CCC” are classified in the class CL01.

(About control unit 130)
The control unit 130 is a controller, and is, for example, various programs (as an example of a classification program) stored in a storage device inside the classification device 100 by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like. (Equivalent) is realized by executing RAM as a work area. Further, the control unit 130 is a controller, and is realized by, for example, an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

As shown in FIG. 3, the control unit 130 according to the embodiment includes an acquisition unit 131, a calculation unit 132, an extraction unit 133, a generation unit 134, a reception unit 135, and a classification unit 136. Realize or execute the functions and actions of information processing described in. The internal configuration of the control unit 130 is not limited to the configuration shown in FIG. 3, and may be any other configuration as long as it is configured to perform information processing described later. Further, the connection relationship of each processing unit included in the control unit 130 is not limited to the connection relationship shown in FIG. 3, and may be another connection relationship.

(About acquisition unit 131)
The acquisition unit 131 acquires various types of information. For example, the acquisition unit 131 acquires a query sent from the user. Specifically, the acquisition unit 131 acquires a keyword arbitrarily input by the user and used for a search process on a search site or the like as a query. The query may include a plurality of keywords.

Further, the acquisition unit 131 acquires the number of users who have input one of two different queries and the number of users who have input both of the two queries in a predetermined aggregation period. do. For example, the acquisition unit 131 is based on information that identifies the user who searched for the first query (user ID in the service, terminal ID that is identification information unique to the terminal, cookie of the browser software, etc.). It is possible to determine whether or not the user has searched for the second query.

In addition, the acquisition unit 131 acquires the total number of searches and the total number of unique users in a predetermined aggregation period. The acquisition unit 131 may acquire the above information from one search site, or may add up the information acquired from a plurality of search sites. In this case, the acquisition unit 131 may acquire information about the user's search behavior from a predetermined external server that provides the search site to the user. Further, the acquisition unit 131 may acquire a query transmitted from a user not only on a search site but also on a predetermined service site (for example, a shopping site or an auction site).

The acquisition unit 131 stores the acquired information in the storage unit 120. Further, the acquisition unit 131 may appropriately acquire the information required by each processing unit, which will be described later, from the storage unit 120.

(About calculation unit 132)
The calculation unit 132 calculates the degree of relevance between the queries acquired by the acquisition unit 131. The calculation unit 132 includes arbitrary queries based on at least the number of users who input one of two different queries and the number of users who input both of the two queries. Calculate the degree of relevance of. When the query includes a plurality of keywords, the calculation unit 132 may count the number of users who have input the keywords for each keyword.

Further, the calculation unit 132 may calculate the degree of relevance based on the number of users who have input two queries that are different from each other within a predetermined period. For example, when the aggregation period is set, the calculation unit 132 calculates the relevance degree based on the number of users who have input both different queries during the aggregation period. In this way, the calculation unit 132 calculates the degree of relevance that captures a relatively long period (such as one year or more) by making the period for aggregating the information that two queries have been searched variable. It is possible to control whether to calculate the degree of relevance that captures a relatively short period.

The calculation unit 132 may classify the users who input the query and calculate the degree of relevance for each of the classified groups. For example, the calculation unit 132 classifies users into predetermined groups based on at least one of a predetermined query input history, a predetermined service usage history, or a predetermined attribute, and is associated with each classified group. The degree may be calculated. For example, the calculation unit 132 may calculate the degree of relevance between queries in a group of users who have a history of searching for queries related to life stages such as "pregnancy" and "childbirth" a predetermined number of times or more. Alternatively, the calculation unit 132 may calculate the degree of relevance between the queries in the user group having a history of using the shopping service or the auction service more than a predetermined number of times, or the user group having a common age group and gender. As a result, the calculation unit 132 can calculate the relevance of the query based not only on the search behavior of an unspecified number of users but also on the search behavior of users in a group having a specific interest.

The calculation unit 132 calculates the degree of relevance between the first query and the second query, which are two queries different from each other, by substituting the numerical values corresponding to the equation shown in the above equation (1). The calculation unit 132 stores the pair of the first query and the second query in the relevance information storage unit 121 in association with the calculated relevance.

(About extraction unit 133)
The extraction unit 133 extracts a plurality of second queries related to the first query based on the degree of relevance between arbitrary queries. For example, the extraction unit 133 extracts a plurality of second queries related to the first query based on the degree of relevance calculated by the calculation unit 132.

Specifically, the extraction unit 133 extracts the second query having a degree of relevance exceeding a predetermined threshold value from the second query with respect to the first query. The predetermined threshold value may be arbitrarily set by, for example, the administrator of the classification device 100, or may be calculated by a statistical method (for example, the average value in all the calculated relevance degrees is predetermined. Set as a threshold, etc.). Further, the extraction unit 133 may extract a predetermined number of second queries from the second query with respect to the first query in descending order of relevance.

(About the generator 134)
The generation unit 134 generates feature information that characterizes the first query based on the plurality of second queries extracted by the extraction unit 133.

For example, the generation unit 134 generates the feature information of the first query related to the plurality of second queries based on the elements constituting each of the plurality of second queries and the number of occurrences of the elements. Specifically, the generation unit 134 may generate information on a combination of the keywords constituting the second query and the number of occurrences of the keywords as the feature information of the first query.

Further, the generation unit 134 may generate a vector in which the elements constituting each of the plurality of second queries are dimensions and the number of appearances of the elements is the number of dimensions of each dimension as the feature information of the first query. .. Specifically, the generation unit 134 uses the keywords constituting each of the extracted second queries as dimensions as the feature information of the first query, and sets the number of occurrences of each keyword as the number of dimensions in each dimension. Generate a word vector to do.

The generation unit 134 associates the first query with the generated feature information and stores it in the feature information storage unit 122.

(About reception desk 135)
The reception unit 135 receives various requests. For example, the reception unit 135 receives an arbitrary keyword from a predetermined user. Specifically, the reception unit 135 accepts an arbitrary keyword from a user who desires the classification of the keyword, and also receives a request (request) for the classification of the keyword. The reception unit 135 sends the received keyword to the classification unit 136. For example, when one keyword is received by the reception unit 135, the classification unit 136 determines whether the one keyword is classified into any of the existing classes.

Further, the reception unit 135 may accept a keyword list including any plurality of keywords. In this case, the reception unit 135 sends the received keyword list to the classification unit 136. In this case, the classification unit 136 classifies each keyword included in the keyword list.

(About classification unit 136)
The classification unit 136 classifies the keywords corresponding to the first query based on the feature information generated by the generation unit 134.

For example, when the feature information of the first query is a vector, the classification unit 136 classifies the keywords based on the similarity (for example, cosine similarity) of the vector generated by the generation unit 134.

Further, when an arbitrary keyword is received by the reception unit 135, the classification unit 136 classifies the keyword based on the feature information corresponding to the keyword. For example, in the classification unit 136, when another keyword belonging to the existing class exists, the accepted keyword is based on the similarity between the feature information of the other keyword and the feature information of the accepted keyword. Determine which class it belongs to.

Further, when the reception unit 135 receives a keyword list including an arbitrary plurality of keywords, the classification unit 136 includes each of the keyword lists included in the keyword list based on the characteristic information of each keyword included in the keyword list. Categorize the keywords of. As a result, the classification unit 136 can appropriately classify the plurality of keywords that the user wants to classify, regardless of the meaning of the keywords, the category, and the like.

The classification unit 136 associates the keyword with the classified class and stores the keyword in the classification information storage unit 123. Further, the classification unit 136 transmits the classification result to the user.

Further, when the classification unit 136 classifies the keyword list received from the user, the classification unit 136 may provide the user with a graph or the like that visualizes the classification result. For example, the classification unit 136 provides the user with information that allows the user to grasp the relationship between the keywords at a glance, such as showing the keywords in a distribution map or the like and showing the keywords classified in the same class (group) in the same color. You may.

[4. Processing procedure]
Next, the procedure of processing by the classification device 100 according to the embodiment will be described with reference to FIGS. 7 and 8. First, the procedure for generating the feature information according to the embodiment will be described with reference to FIG. 7. FIG. 7 is a flowchart (1) showing a processing procedure by the classification device 100 according to the embodiment.

As shown in FIG. 7, the classification device 100 determines whether or not the query transmitted from the user has been acquired (step S101). If the query has not been acquired (step S101; No), the classification device 100 waits until the query is acquired.

On the other hand, when a query is acquired (step S101; Yes), the classification device 100 selects an arbitrary first query to be processed from the acquired queries (step S102). Further, the classification device 100 aggregates the second query searched by the user who searched for the first query (step S103).

Then, the classification device 100 calculates the degree of association between the first query and the second query by using, for example, the above equation (1) (step S104). After that, the classification device 100 determines whether or not the relevance of all search queries has been calculated (step S105). When the relevance of all queries has not been calculated (step S105; No), the classification device 100 repeats the processes of steps S102 to S104.

On the other hand, when the relevance of all the queries is calculated (step S105; Yes), the classification device 100 extracts a plurality of second queries having a relevance exceeding a predetermined threshold value for any first query (step S106). ). Then, the classification device 100 analyzes the extracted second query by morphological analysis (step S107).

Subsequently, the classification device 100 generates the feature information of the first query based on the morpheme and the number of appearances of each morpheme (step S108). Then, the classification device 100 stores the generated feature information in the storage unit 120 (step S109).

Next, the procedure of the classification process according to the embodiment will be described with reference to FIG. FIG. 8 is a flowchart (2) showing a processing procedure by the classification device 100 according to the embodiment.

As shown in FIG. 8, the classification device 100 determines whether or not the keyword list has been accepted from the user (step S201). If the keyword list is not accepted (step S201; No), the classification device 100 waits until it is accepted.

On the other hand, when the keyword list is received (step S201; Yes), the classification device 100 refers to, for example, the feature information storage unit 122, and identifies the feature information corresponding to each keyword included in the keyword list (step S202). ).

Then, the classification device 100 calculates the degree of similarity between the feature information of each keyword (step S203). Further, the classification device 100 classifies keywords based on the calculated similarity (step S204). Then, the classification device 100 stores the classification result in the storage unit 120 (step S205). Further, the classification device 100 transmits the classification result to the user who is the transmission source of the keyword list (step S206).

[5. Modification example]
The classification system 1 according to the above-described embodiment may be implemented in various different forms other than the above-mentioned embodiment. Therefore, in the following, other embodiments of each device included in the above classification system 1 will be described.

[5-1. Keyword list generation]
In the above embodiment, the classification device 100 has shown an example of accepting a keyword list from a user. Here, the classification device 100 may generate a keyword list based on one keyword received from the user, and may perform a process of classifying the keywords included in the generated keyword list.

For example, when the classification device 100 receives the input of an arbitrary keyword from a predetermined user, the classification device 100 extracts a plurality of keywords whose relevance to the arbitrary keyword exceeds a predetermined threshold value, and extracts the arbitrary keyword. Generate a keyword list that contains multiple keywords. Then, the classification device 100 classifies each keyword included in the generated keyword list. This point will be described with reference to FIG. 9 along with the flow of processing. FIG. 9 is a flowchart showing a processing procedure by the classification device according to the modified example.

As shown in FIG. 9, the classification device 100 determines whether or not the keyword has been accepted from the user (step S301). If the keyword is not accepted (step S301; No), the classification device 100 waits until it is accepted.

On the other hand, when the keyword is received (step S301; Yes), the classification device 100 refers to, for example, the relevance information storage unit 121, and extracts a keyword having a relevance degree exceeding a predetermined threshold value with respect to the accepted keyword. (Step S302).

Then, the classification device 100 generates a keyword list including the received keywords and the keywords extracted in step S302 (step S303). Then, the classification device 100 executes a classification process for the generated keyword list, for example, according to the flow shown in FIG. 8 (step S304).

In this way, the classification device 100 may generate a keyword list based on the keywords received from the user, and may perform a process of classifying the keywords included in the generated keyword list. As a result, the user does not need to create a list by himself / herself, and by sending only one keyword of interest to the classification device 100, the classification result regarding the keyword group related to the keyword can be obtained. can. That is, the user can obtain a group of keywords that have been grouped to some extent with respect to one input keyword. Therefore, if you are a marketing user, for example, what kind of needs do you have regarding one entered keyword, and what other keywords are interested in a user who is interested in one entered keyword? It is possible to obtain information such as whether or not it is possessed.

[5-2. Morphological analysis for queries]
In the above embodiment, the classification device 100 shows an example of calculating the degree of relevance between queries based on the number of users who have input both the first query and the second query. Here, as described above, the query may be composed of not only one keyword but also a plurality of keywords and sentences. Therefore, the classification device 100 may perform morphological analysis of the query sent from the user and treat the keywords included in the result of the morphological analysis as the first query or the second query. In this case, the classification device 100 may extract nouns and proper nouns included in the query using a known description, and use only the extracted keywords for processing.

[5-3. Search behavior]
The classification device 100 does not necessarily have to be conditional on the search action being performed in the same period as the aggregation period for the period for determining both the first query and the second query as the searched user. That is, the classification device 100 may set a period for determining both the first query and the second query as the searched user and a period for totaling the number of users who searched the query. For example, the classification device 100 may treat the user as a user who has searched for both the first query and the second query when the first query and the second query are sent from the same user within 24 hours. good. Further, the classification device 100 is used when the first query and the second query are transmitted in the same session (for example, a series of actions from accessing a predetermined search site until the access is interrupted) by the same user. May be treated as a user who searched for both the first query and the second query. In this way, the classification device 100 may flexibly handle the user's search behavior and perform various information processing.

[5-4. Search site]
In the above embodiment, an example is shown in which the search site is provided by the classification device 100. However, the search site may be provided by a predetermined external server (eg, a web server that provides a search service). In this case, the classification device 100 may acquire the query, the user information, and the like transmitted by the user to the search site via the external server.

[6. others〕
Further, among the processes described in the above-described embodiment, all or a part of the processes described as being automatically performed can be manually performed, or the processes described as being manually performed can be performed. All or part of it can be done automatically by a known method. In addition, information including processing procedures, specific names, various data and parameters shown in the above documents and drawings can be arbitrarily changed unless otherwise specified.

Further, each component of each of the illustrated devices is a functional concept, and does not necessarily have to be physically configured as shown in the figure. That is, the specific form of distribution / integration of each device is not limited to the one shown in the figure, and all or part of them may be functionally or physically distributed / physically in any unit according to various loads and usage conditions. Can be integrated and configured.

For example, the relevance information storage unit 121, the feature information storage unit 122, and the classification information storage unit 123 shown in FIG. 3 may not be held by the classification device 100 but may be held by an external storage server or the like. In this case, the classification device 100 acquires relevance information, feature information, and the like by accessing the storage server.

Further, for example, the classification device 100 described above has a front-end server side that mainly executes communication with an external device, such as acquiring a query from a user terminal 10 and transmitting a classification result to a user, and characteristic information. It may be distributed to the back-end server side that executes the process of classifying queries based on.

[7. Hardware configuration]
Further, the classification device 100 and the user terminal 10 according to the above-described embodiment are realized by, for example, a computer 1000 having a configuration as shown in FIG. Hereinafter, the classification device 100 will be described as an example. FIG. 10 is a hardware configuration diagram showing an example of a computer 1000 that realizes the functions of the classification device 100. The computer 1000 includes a CPU 1100, a RAM 1200, a ROM (Read Only Memory) 1300, an HDD (Hard Disk Drive) 1400, a communication interface (I / F) 1500, an input / output interface (I / F) 1600, and a media interface (I / F). ) Has 1700.

The CPU 1100 operates based on a program stored in the ROM 1300 or the HDD 1400, and controls each part. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 is started, a program depending on the hardware of the computer 1000, and the like.

The HDD 1400 stores a program executed by the CPU 1100, data used by such a program, and the like. The communication interface 1500 receives data from another device via the communication network 500 (corresponding to the network N shown in FIG. 2) and sends the data to the CPU 1100, and also receives data generated by the CPU 1100 via the communication network 500. Send to the device of.

The CPU 1100 controls an output device such as a display or a printer, and an input device such as a keyboard or a mouse via the input / output interface 1600. The CPU 1100 acquires data from the input device via the input / output interface 1600. Further, the CPU 1100 outputs the data generated via the input / output interface 1600 to the output device.

The media interface 1700 reads a program or data stored in the recording medium 1800 and provides the program or data to the CPU 1100 via the RAM 1200. The CPU 1100 loads the program from the recording medium 1800 onto the RAM 1200 via the media interface 1700, and executes the loaded program. The recording medium 1800 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 MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. And so on.

For example, when the computer 1000 functions as the classification device 100, the CPU 1100 of the computer 1000 realizes the function of the control unit 130 by executing the program loaded on the RAM 1200. Further, each data in the storage unit 120 is stored in the HDD 1400. The CPU 1100 of the computer 1000 reads these programs from the recording medium 1800 and executes them, but as another example, these programs may be acquired from another device via the communication network 500.

[8. effect〕
As described above, the classification device 100 according to the embodiment includes an extraction unit 133, a generation unit 134, and a classification unit 136. The extraction unit 133 extracts a plurality of second queries related to the first query based on the degree of relevance between arbitrary queries. The generation unit 134 generates feature information that characterizes the first query based on the plurality of second queries extracted by the extraction unit 133. The classification unit 136 classifies the keywords corresponding to the first query based on the feature information generated by the generation unit 134.

As described above, the classification device 100 according to the embodiment does not depend on qualitative information such as the meaning and properties of the query transmitted from the user, but classifies based on the quantitative information of the actual search behavior of the user. I do. In other words, the classification device 100 classifies related keywords without categorizing each query. For example, according to the classification device 100, keywords are classified by reflecting the user's hobbies, tastes, and interests, so that even keywords that do not seem to be directly related in meaning belong to the same group (class). It can be classified as a keyword. In this way, the classification device 100 can perform classification for various queries transmitted from the user, and as a result, it is possible to derive mutual relationships between various queries.

Further, the generation unit 134 generates the feature information of the first query related to the plurality of second queries based on the elements constituting each of the plurality of second queries and the number of occurrences of the elements.

As described above, the classification device 100 according to the embodiment shows the characteristics of the first query by the elements constituting the second query (for example, the keywords included in the second query) and the number of occurrences of the elements. You can capture the features in detail. Therefore, the classification device 100 can appropriately classify the keywords corresponding to the first query.

Further, the generation unit 134 generates a vector in which the element constituting each of the plurality of second queries is set as the dimension and the number of appearances of the element is set as the number of dimensions of each dimension as the feature information of the first query. The classification unit 136 classifies keywords based on the similarity of the vectors generated by the generation unit 134.

As described above, the classification device 100 according to the embodiment can easily calculate the similarity between the feature information by expressing the element constituting the second query and the number of appearances of the element by a vector.

Further, the classification device 100 according to the embodiment is at least based on the number of users who input one of two different queries among arbitrary queries and the number of users who input both of the two queries. Further, a calculation unit 132 for calculating the degree of relevance between arbitrary queries is provided. The extraction unit 133 extracts a plurality of second queries related to the first query based on the degree of relevance calculated by the calculation unit 132.

As described above, since the classification device 100 according to the embodiment calculates the relevance degree based on the user's search behavior log, the relevance degree is based on the actual user's interests regardless of the meaning or category of the query itself. Can be calculated. As a result, the classification device 100 can calculate the degree of relevance that accurately reflects the user's behavior, such as calculating the degree of relevance between the queries searched by the users who have similar behaviors.

Further, the calculation unit 132 calculates the degree of relevance based on the number of users who have input two queries that are different from each other within a predetermined period.

In this way, the classification device 100 according to the embodiment sets a predetermined period, captures a series of user search behaviors in the predetermined period, and calculates the degree of relevance. The classification device 100 can extract queries related to changes in the user's life stage, such as pregnancy, childbirth, and baby products, as related queries by setting a predetermined period in a span of several years, for example. .. As a result, the classification device 100 can perform classification processing that reflects a series of user actions.

Further, the classification device 100 according to the embodiment further includes a reception unit 135 that accepts an arbitrary keyword from a predetermined user. The classification unit 136 classifies the keyword based on the feature information corresponding to the keyword received by the reception unit 135.

As described above, the classification device 100 according to the embodiment classifies the keywords received from the users, so that the keywords desired by the users are searched by the users who are interested in them. , Information such as what other keywords the keyword is related to can be provided to the user.

Further, the reception unit 135 accepts a keyword list including an arbitrary plurality of keywords. The classification unit 136 classifies each keyword included in the keyword list based on the characteristic information of each keyword included in the keyword list received by the reception unit 135.

As described above, the classification device 100 according to the embodiment can provide the user with information such as what kind of keywords tend to be easily searched by the user by classifying the keywords included in the keyword list. ..

Further, when the reception unit 135 receives the input of an arbitrary keyword from a predetermined user, the reception unit 135 extracts a plurality of keywords whose relevance to the arbitrary keyword exceeds a predetermined threshold value, and extracts the arbitrary keyword. Generate a keyword list that contains multiple keywords. The classification unit 136 classifies each keyword included in the keyword list generated by the reception unit 135.

As described above, the classification device 100 according to the embodiment may generate a keyword list based on the keywords received from the user and perform a process of classifying the keywords included in the generated keyword list. As a result, the classification device 100 can provide the user with the classification result regarding the keyword group related to the keyword without causing the user to take the trouble of creating the keyword list.

Although some of the embodiments of the present application have been described in detail with reference to the drawings, these are examples, and various modifications are made based on the knowledge of those skilled in the art, including the embodiments described in the disclosure column of the invention. It is possible to carry out the present invention in other modified forms.

Further, the above-mentioned classification device 100 may be realized by a plurality of server computers, and depending on the function, it may be realized by calling an external platform or the like by API (Application Programming Interface), network computing, or the like. It can be changed flexibly.

Further, the "section, module, unit" described in the claims can be read as "means" or "circuit". For example, the acquisition unit can be read as an acquisition means or an acquisition circuit.

1 Classification system 10 User terminal 100 Classification device 110 Communication unit 120 Storage unit 121 Relevance information storage unit 122 Feature information storage unit 123 Classification information storage unit 130 Control unit 131 Acquisition unit 132 Calculation unit 133 Extraction unit 134 Generation unit 135 Reception unit 136 Classification department

Claims (9)

An extractor that extracts multiple second queries related to the first query based on the degree of relevance between arbitrary queries,
A generation unit that generates feature information that characterizes the first query based on a plurality of second queries extracted by the extraction unit.
A classification unit that classifies keywords corresponding to the first query based on the feature information generated by the generation unit.
Equipped with
The generator is
Based on the elements constituting each of the plurality of second queries and the number of occurrences of the elements, the feature information of the first query related to the plurality of second queries is generated.
A classification device characterized by that. The generator is
As the feature information of the first query, a vector is generated in which the elements constituting each of the plurality of second queries are dimensions and the number of appearances of the elements is the number of dimensions of each dimension.
The classification unit
The keywords are classified based on the similarity of the vectors generated by the generator.
The classification device according to claim 1 , wherein the classification device is characterized by the above. Calculate the relevance of any query based on at least the number of users who entered one of two different queries out of any query and the number of users who entered both of those two queries. Calculation unit,
Further prepare
The extraction unit
A plurality of second queries related to the first query are extracted based on the degree of relevance calculated by the calculation unit.
The classification device according to claim 1 or 2 , wherein the classification device is characterized by the above. The calculation unit
The relevance is calculated based on the number of users who have entered both of the two different queries within a given period of time.
The classification device according to claim 3 , wherein the classification device is characterized by the above. Reception department that accepts arbitrary keywords from a predetermined user,
Further prepare
The classification unit
The keyword is classified based on the feature information corresponding to the keyword received by the reception unit.
The classification device according to any one of claims 1 to 4 , wherein the classification device is characterized by the above. The reception department
Accepts a keyword list containing any number of keywords,
The classification unit
Based on the characteristic information of each keyword included in the keyword list accepted by the reception unit, each keyword included in the keyword list is classified.
The classification device according to claim 5 , wherein the classification device is characterized by the above. The reception department
When the input of an arbitrary keyword is received from the predetermined user, a plurality of keywords whose relevance to the arbitrary keyword exceeds a predetermined threshold value are extracted, and the arbitrary keyword and the extracted plurality of keywords are included. Generate a keyword list and
The classification unit
Classify each keyword included in the keyword list generated by the reception unit.
The classification device according to claim 5 , wherein the classification device is characterized by the above. It ’s a computer-executed classification method.
An extraction process that extracts multiple second queries related to the first query based on the degree of relevance between arbitrary queries.
A generation step that generates feature information that characterizes the first query based on a plurality of second queries extracted by the extraction step.
A classification step of classifying keywords corresponding to the first query based on the feature information generated by the generation step, and a classification step.
Including
The production step is
Based on the elements constituting each of the plurality of second queries and the number of occurrences of the elements, the feature information of the first query related to the plurality of second queries is generated.
A classification method characterized by that. An extraction procedure that extracts multiple second queries related to the first query based on the degree of relevance between arbitrary queries, and
A generation procedure that generates feature information that characterizes the first query based on a plurality of second queries extracted by the extraction procedure.
A classification procedure for classifying keywords corresponding to the first query based on the feature information generated by the generation procedure, and a classification procedure.
Let the computer run
The generation procedure is
Based on the elements constituting each of the plurality of second queries and the number of occurrences of the elements, the feature information of the first query related to the plurality of second queries is generated.
A classification program characterized by that.

Images