JP2017199258A - Related topic detection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To more accurately detect a related topic.SOLUTION: A related topic detection device includes: a section topic information storage part 14 for storing section topic information which associates a topic of the contributed information contributed in the section for each section which is a range of the predetermined time with a user having contributed the contributed information; and a related topic detection part 15 for detecting a second topic of a second section relating to a first topic of a first section based on the user who is shown in the section topic information stored by the section topic information storage part 14 and who contributes the contributed information of the first topic and the second topic. The related topic detection part 15 may perform the detection based on a similarity between the user who contributes the contributed information of the first topic within a first section and the user who contributes the contributed information of the second topic within a second section.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a related topic detection apparatus that detects a topic related to a certain topic in a situation where a topic of posted information posted by a user changes in time series.

  2. Description of the Related Art Conventionally, in a message exchange system, a technique for performing display related to a topic related to the contents of time-series messages exchanged between user terminals is known (Patent Document 1 below).

Japanese Patent Laid-Open No. 2006-2178

  Generally, as a technique for clarifying the relevance (structure) of topical time series changes, there is a technique for comparing the appearance probabilities of words in a topic for each unit time. However, although this method can clarify the relationship between topics having similar appearance probabilities, there is a problem that the relationship between the topics is not clarified when the appearance probabilities of words between topics are different. That is, there is a problem that related topics cannot be reliably detected.

  Therefore, the present invention has been made in view of such problems, and an object of the present invention is to provide a related topic detection apparatus that can more reliably detect related topics.

  In order to solve the above problems, the related topic detection device of the present invention associates, for each section that is a predetermined time range, the topic of posted information posted in the section and the user who posted the posted information. Section topic information storage means for storing section topic information and a second topic of the second section related to the first topic of the first section are indicated in the section topic information stored by the section topic information storage means. Relevant topic detection means for detecting based on a user who has posted information on the first topic and the second topic.

  With this configuration, the second topic (related topic) in the second section related to the first topic (related source topic) in the first section is converted into the first topic and the related topic detection unit by the related topic detection unit. It is detected based on the user who posted the posting information of the second topic. Thereby, for example, even when the appearance probabilities of the words between the first topic and the second topic are different, more reliably based on the user who posted the posting information of the first topic and the second topic, The second topic can be detected as a topic related to the first topic. That is, a related topic can be detected more reliably.

  In the related topic detection apparatus of the present invention, the related topic detection means includes a user who posted the first topic posting information in the first section, and the second topic posting information in the second section. It is good also as detecting based on the similarity with the user who contributed. Generally, the posting information of each of two related topics is characterized by a high possibility that a user with a high degree of similarity (more similar) has posted. With this configuration, using the feature, a user who posted the posting information on the first topic in the first section, and a user who posted the posting information on the second topic in the second section, Based on the similarity, the second topic related to the first topic can be detected more reliably.

  In the related topic detection apparatus of the present invention, the related topic detection means posts the posting information of the first topic within the first section, and the posting information of the second topic within the second section. It may be detected based on the number of posted users. Generally, the posting information of each of two related topics is characterized by the high possibility that the same user has posted. If such a configuration is adopted, the posting information of the first topic is posted in the first section and the posting information of the second topic is posted in the second section using the feature. Based on the number, the second topic related to the first topic can be detected more reliably.

  The related topic detection device of the present invention stores the second topic in the second section related to the first topic in the first section by the related topic detection means when the second topic is detected by the section topic information storage means. Out of the users who posted the posting information of the first topic in the first section, the users who have not posted the posting information of the second topic in the second section are extracted based on the section topic information Then, it may be further provided with topic output means for performing output based on the second topic for the extracted user. With this configuration, when the related topic detection unit detects the second topic related to the first topic, the user who posted the posting information of the first topic within the first section by the topic output unit Among them, the user who has not posted the posting information of the second topic within the second section is extracted, and the output based on the second topic is performed on the extracted user. Thereby, for example, among the users who have made postings related to the first topic, the user who is not aware of the existence of the second topic related to the first topic is aware of the existence of the second topic. Can be made. That is, it is possible to improve the convenience of the user, such as making the user aware of a topic that is related to a topic that the user is interested in and that the user is not aware of.

  Further, the related topic detection apparatus of the present invention detects the second topic in the second section related to the first topic in the first section and the second topic when the related topic detection unit detects the second topic in the second section. The degree of relevance to the topic is calculated based on the user who is indicated in the section topic information stored by the section topic information storage means and has posted the posting information of the first topic and the second topic The degree calculation means may be further provided, and the topic output means may perform output based on the second topic based on the degree of association calculated by the degree of association calculation means. With this configuration, when the second topic related to the first topic is detected by the related topic detecting unit, the degree of association between the first topic and the second topic is calculated by the related level calculating unit. It is calculated based on the user who posted the posting information of the first topic and the second topic, and the topic output means performs output based on the second topic based on the calculated degree of association. Thereby, since the output based on the second topic related to the first topic is performed based on the degree of association between the first topic and the second topic, for example, a topic having a higher degree of association is given priority. The convenience of the user can be improved, such as being able to display to the user.

  The related topic detection apparatus of the present invention further includes user information storage means for storing user information in which a user is associated with the degree of influence of the user on other users, and the related degree calculation means is user information storage means. , The posting information of the first topic is posted within the first section, and the posting information of the second topic is calculated based on the degree of influence associated with the user who posted the posting information within the second section It is good as well. In general, the degree of relevance between two topics is characterized in that it is related to the degree of influence of a user who has posted on the two topics with respect to other users. By adopting such a configuration, it is possible to use the feature to post the posting information on the first topic in the first section and to the user who posted the posting information on the second topic in the second section. The degree of association between the first topic and the second topic can be calculated more reliably based on the associated degree of influence.

  According to the present invention, related topics can be detected more reliably.

It is a functional block diagram of the related topic detection apparatus which concerns on embodiment of this invention. It is a hardware block diagram of the related topic detection apparatus which concerns on embodiment of this invention. It is a figure which shows the example of a table of contribution information history. It is a figure which shows the example of a table of user information. It is a figure which shows the example of a table of area topic information. It is a figure which shows the example of a table of related topic information. It is a figure which shows the example of a table of output topic information. It is a flowchart which shows the process performed with the related topic detection apparatus which concerns on embodiment of this invention. It is a figure which shows the comparison with the technique by the related topic detection apparatus which concerns on the related technique detection method which concerns on embodiment of this invention.

  Hereinafter, embodiments of a related topic detection apparatus according to the present invention will be described in detail with reference to the drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted.

  FIG. 1 is a functional block diagram of a related topic detection apparatus 1 according to an embodiment of the present invention. As shown in FIG. 1, the related topic detection device 1 includes a posting information history storage unit 10, a user information storage unit 11 (user information storage unit), an instruction acquisition unit 12, a section topic calculation unit 13, and a section topic information storage unit 14 ( Section topic information storage unit), related topic detection unit 15 (related topic detection unit), related topic information storage unit 16, relevance degree calculation unit 17 (relevance degree calculation unit), topic output unit 18 (topic output unit), and output topic An information storage unit 19 is included.

  FIG. 2 is a hardware configuration diagram of the related topic detection apparatus 1. As shown in FIG. 2, the related topic detection apparatus 1 shown in FIG. 1 physically includes one or a plurality of CPUs 100, a RAM 101 and a ROM 102 as main storage devices, and an input / output device 103 such as a keyboard, a mouse, and a display. The computer system includes a communication module 104 which is a data transmission / reception device such as a network card, and an auxiliary storage device 105 such as a hard disk or a semiconductor memory.

  Each function in FIG. 1 causes the input / output device 103 and the communication module 104 to operate under the control of the CPU 100 by reading predetermined computer software on the hardware such as the CPU 100 and the RAM 101 shown in FIG. This is realized by reading and writing data in the RAM 101 and the auxiliary storage device 105.

  Instead of executing the functions shown in FIG. 1 by a processor such as the CPU 100, all or a part of the functions may be executed by constructing a dedicated integrated circuit. For example, the above functions may be executed by constructing a dedicated integrated circuit for performing image processing and communication control.

  Hereinafter, each functional block of the related topic detection apparatus 1 illustrated in FIG. 1 will be described.

  The posted information history storage unit 10 stores a posted information history that is a time-series history of posted information posted by the user. Posted information is content such as text and images. Specific examples include text (tweets) that expresses the user's situation and notes, emails created by the user, images captured by the user, and users on the Internet. Product purchase data related to products purchased at a shopping site. The posted information history is information in which posted information is associated with time information such as time and date when the posted information is posted (or created). In the present embodiment, it is assumed that the related topic detection device 1 has a microblogging function, and tweets posted by a plurality of users are accumulated in the posting information history storage unit 10.

  FIG. 3 is a diagram illustrating a table example of the posted information history stored by the posted information history storage unit 10. As shown in the table example in FIG. 3, the posted information history includes a “post ID” indicating identification information of each posted information, a “posted content” indicating the content (content) of the posted information, and the user who posted the posted information. “Posting user” indicating user identification information as identification information and “posting date” indicating posting date (posting time) when the posting information was posted by the user are associated with each other.

  In the present embodiment, the related topic detection device 1 includes the posted information history storage unit 10, but an external server device connected to the related topic detection device 1 through a network is configured with the same function as the posted information history storage unit 10 ( ) May be provided. In that case, the related topic detection device 1 can realize the same processing as the processing described in the present embodiment by transmitting / receiving necessary information and the like to / from the posted information history storage unit 10 via the network. it can.

  The user information storage unit 11 stores user information related to the user, such as indicating the status of the user. Specifically, the user information storage unit 11 stores user information in which a user is associated with the degree of influence of the user on other users. More specifically, the user information includes the number of subscribers (the number of followers) and the number of subscribers (the number of followers) of the users who post the posted information. The number of subscribers is the number of users who subscribe (follow) the posting of the target user. Note that the subscription includes a subscription (application, contract, registration) in which no money exchange occurs. The number of subscribers is the number of users subscribed (followed) by the target user. FIG. 4 is a diagram illustrating an example of a table of user information stored by the user information storage unit 11. As shown in the table example in FIG. 4, the user information includes “user ID” indicating user identification information, “number of followers” indicating the number of subscribers of the user indicated by the user identification information, and user information indicated by the user identification information. “Number of followers” indicating the number of subscribers is associated.

  The instruction acquisition unit 12 acquires instruction information related to an instruction from an administrator or the like of the related topic detection device 1. The instruction information includes information for instructing the start of detection of related topics. The instruction information may include other information, and what information may be included will be described later as appropriate. The instruction acquisition unit 12 may acquire the instruction information from another device via the network and communication module 104, or may acquire it via the input / output device 103 (for example, a keyboard). The instruction acquisition unit 12 outputs the acquired instruction information to the section topic calculation unit 13, the related topic detection unit 15, the relevance calculation unit 17, the topic output unit 18, and the like. The instruction acquisition unit 12 may also temporarily store the acquired instruction information in the related topic detection apparatus 1 so that other functional blocks can refer to and use the information appropriately.

  When the instruction information is input from the instruction acquisition unit 12, the section topic calculation unit 13 is based on the posting information history stored by the posting information history storage unit 10, in each section within a predetermined time range. Section topic information indicating the topic of the posted information posted and the user who posted the posted information is calculated. This will be specifically described below. When the instruction information for instructing the start of the detection of the related topic is input from the instruction acquisition unit 12, the section topic calculation unit 13 selects an arbitrary plurality of post information histories stored in the post information history storage unit 10. The post information history of the sample user who is the user is extracted. The sample users may be randomly determined by the section topic calculation unit 13, or a predetermined number of users (for example, 100) in descending order of the number of users having a large number of posts in a predetermined period (for example, the past three months). A user included in the person) or all users included in the posted information history. Further, the posted information history to be extracted may be for the entire period of the posted information history stored by the posted information history storage unit 10 or for a predetermined period (for example, the past one month). Here, the information regarding the above-mentioned “predetermined period”, “predetermined number of people”, and “predetermined period” used by the section topic calculating unit 13 may be stored in the related topic detecting device 1 in advance, It may be included in the instruction information input from the instruction acquisition unit 12.

Next, the section topic calculation unit 13 divides the extracted posted information history into sections that are within a predetermined time range (for example, one week). Here, the information regarding the above-mentioned “section” used by the section topic calculation unit 13 may be stored in advance in the related topic detection device 1, for example, or included in the instruction information input from the instruction acquisition unit 12. May be. Next, the section topic calculation unit 13 performs topic modeling, which is a conventional technique, on the posted information included in the post information history for each divided section, thereby posting posts posted in the section for each section. Section topic information indicating the topic of information and the user who posted the posted information is calculated. One example of topic modeling is LDA (Latent Dirichlet Allocation) disclosed in the following non-patent literature. For the specific calculation method of LDA, refer to the following non-patent literature.
Non-patent literature: Blei, David M., Andrew Y. Ng, and Michael I. Jordan. "Latent dirichlet allocation." The Journal of machine Learning research 3 (2003): 993-1022.

  FIG. 5 is a diagram illustrating a table example of the section topic information calculated by the section topic calculation unit 13. As shown in the example of the table in FIG. 5, the section topic information includes “section ID” indicating section identification information that is section identification information, and topic identification information that is topic identification information of posted information posted in the section. "Topic ID" to indicate, "Word Appearance Probability Distribution" indicating the probability distribution of words constituting the topic, and user identification information of (one or more) users who posted posting information related to the topic in the section “Posting user” is associated. The section topic calculation unit 13 outputs (stores) the calculated section topic information to the section topic information storage unit 14 and also stores section topic information calculation completion information indicating that the calculation of section topic information has been completed. Output to.

  Note that the timing at which the section topic calculation unit 13 calculates the section topic information is not limited to the case where the instruction information is input from the instruction acquisition unit 12, but periodically (for example, once a week) in the background. Also good.

  The section topic information storage unit 14 stores the section topic information input from the section topic calculation unit 13. That is, the section topic information storage unit 14 identifies, for each section within a predetermined time range, the topic of the posted information posted in the section (topic identification information for identifying) and the user who posted the posted information ( Section topic information associated with user identification information). For example, the section topic information storage unit 14 stores the section topic information shown in the table example of FIG.

  The related topic detection unit 15 is a user indicated in the section topic information stored in the section topic information storage unit 14 by referring to the second topic of the second section related to the first topic of the first section. , Based on the user who posted the posting information of the first topic and the second topic. The related topic detection unit 15 is based on the similarity between the user who posted the first topic posting information in the first section and the user who posted the second topic posting information in the second section. It may be detected. Further, the related topic detection unit 15 posts the posting information of the first topic within the first section, and based on the number of users who posted the posting information of the second topic within the second section. It may be detected. This will be specifically described below.

First, when the topic topic information calculation completion information is input from the section topic calculation unit 13, the related topic detection unit 15 first of the first section among the section topic information stored by the section topic information storage unit 14. Topic and the second topic of the second interval. The first section may be any section of the section topic information stored by the section topic information storage unit 14 or may be a section specified in the instruction information acquired by the instruction acquisition unit 12. Alternatively, it may be a previous section in time series of a section including the current time (obtained by using a built-in function included in the related topic detection apparatus 1). Hereinafter, the first section is referred to as section t. The first topic may be an arbitrary topic included in the section t, or may be a topic specified in the instruction information acquired by the instruction acquisition unit 12. Hereinafter, the first topic is referred to as Topic _i . The second section may be a first order section in time series of the section t, or may be an arbitrary section of the section topic information stored by the section topic information storage unit 14, It may be a section specified in the instruction information acquired by the instruction acquisition unit 12, or may be a section including the current time (acquired using a built-in function included in the related topic detection device 1). . Hereinafter, the second section is referred to as section t + 1. The second topic may be an arbitrary topic included in the section t + 1, or may be a topic specified in the instruction information acquired by the instruction acquisition unit 12. Hereinafter, the second topic is denoted as Topic _j .

Next, the related topic detection unit 15 refers to the section topic information stored by the section topic information storage unit 14, and obtains the user identification information of the user who posted the posting information of Topic _i of the selected section t in the section t. get. Hereinafter, a set of users indicated by the acquired user identification information is referred to as U _i . For example, in the example of the table illustrated in FIG. 5, when the section identification information of the section t selected by the related topic detection unit 15 is “I001” and the topic identification information of the selected Topic _i is “T001”, the related topic detection unit 15 acquires “U001”, “U003”, and “U005”, which are user identification information associated with the record including the section identification information and the topic identification information. Similarly, the related topic detection unit 15 refers to the section topic information stored by the section topic information storage unit 14, and obtains the user identification information of the user who posted the posted information of Topic _j of the selected section t + 1 in the section t + 1. get. Hereinafter, a set of users indicated by the acquired user identification information is denoted as U _j .

ここで、｜Ｕ_ｉ｜は集合Ｕ_ｉに含まれるユーザの人数を示し、｜Ｕ_ｊ｜は集合Ｕ_ｊに含まれるユーザの人数を示し、｜Ｕ_ｉ∩Ｕ_ｊ｜は集合Ｕ_ｉと集合Ｕ_ｊとの共通部分に含まれるユーザの人数を示す。なお、｜Ｕ_ｉ∩Ｕ_ｊ｜は、Ｔｏｐｉｃ_ｉの投稿情報を区間ｔ内に投稿し、かつ、Ｔｏｐｉｃ_ｊの投稿情報を区間ｔ＋１内に投稿したユーザの数である。次に、関連トピック検出部１５は、算出されたｓｉｍ（Ｕ_ｉ，Ｕ_ｊ）が所定の閾値以上であるか否かを判定し、閾値以上であればＴｏｐｉｃ_ｊ（関連先トピック）がＴｏｐｉｃ_ｉ（関連元トピック）に関連すると判定し（Ｔｏｐｉｃ_ｉに関連するトピックとしてＴｏｐｉｃ_ｊを検出し）、閾値未満であればＴｏｐｉｃ_ｊがＴｏｐｉｃ_ｉに関連しないと判定する。所定の閾値は、例えば関連トピック検出装置１に予め格納されているものとし、関連トピック検出部１５は当該所定の閾値を参照及び利用する。 Next, the related topic detection unit 15 determines whether Topic _j is related to Topic _i based on U _i and U _j . For example, the related topic detection unit 15 may make a determination based on comparison using a dice coefficient. In this case, related topics detector 15, first, _{U i} and _{U j} and the similarity sim _(U i, _{U j)} is calculated by equation (1) below.

Here, | U _i | indicates the number of users included in the set U _i , | U _j | indicates the number of users included in the set U _j , and | U _i ∩U _j | indicates the set U _i and the set U _i The number of users included in the common part with U _j is shown. Note that | U _i ∩U _j | is the number of users who posted Topic _i posting information in section t and posted Topic _j posting information in section t + 1. Next, Related Topics detector 15 calculated sim _(U i, U _j) is equal to or greater than a predetermined threshold value, if the above threshold Topic in _j (associated destination topics) is Topic in _i (Topic _j is detected as a topic related to Topic _i ), and if it is less than the threshold, it is determined that Topic _j is not related to Topic _i . For example, it is assumed that the predetermined threshold value is stored in advance in the related topic detection device 1, and the related topic detection unit 15 refers to and uses the predetermined threshold value.

Next, when the related topic detection unit 15 determines that Topic _j is related to Topic _i , the related topic detection unit 15 calculates the determination result as related topic information. Related Topics information indicates that it has detected a Topic in _j associated with Topic in _i, including a set of the topic identification information topic identification information and Topic in _j the Topic in _i. The related topic detection unit 15 may detect one or more second topics in one or more second sections, which are related to one or more first topics in one or more first sections. Good. For example, the related topic detection unit 15 performs, for each of all topics included in the first section specified in the instruction information acquired by the instruction acquisition unit 12, the chronological order of the first section. Secondly, it may be determined whether or not all topics included in the section are related, and all related topics may be detected. In that case, the related topic detection unit 15 calculates related topic information that is a set of identification information of the related source topic detected as related and identification information of the related destination topic.

FIG. 6A is a diagram illustrating a table example of related topic information calculated by the related topic detection unit 15. As shown in the table example of FIG. 6A, the related topic information is associated with “related source topic” indicating the topic identification information of the related source topic and “related destination topic” indicating the topic identification information of the related destination topic. It has been. The related topic detection unit 15 outputs (stores) the calculated related topic information to the related topic information storage unit 16 and also stores related topic information calculation completion information indicating that the output of the related topic information is completed. And output to the topic output unit 18. Note that the related topic detection unit 15 may output the calculated related topic information to the relevance degree calculation unit 17 and the topic output unit 18. If the related topic detection unit 15 determines that Topic _j is not related to Topic _i , the related topic detection unit 15 may output the determination result (that Topic _j has been detected not related to Topic _i ) to the topic output unit 18. Good.

  The related topic information storage unit 16 stores the related topic information output from the related topic detection unit 15. For example, the related topic information storage unit 16 stores the related topic information shown in the table example of FIG.

When the related topic detection unit 15 detects the Topic _j of the section t + 1 related to the Topic _i of the section t, the related degree calculation unit 17 determines the relevance between the Topic _i and Topic _j by the section topic information storage unit 14. The calculation is based on the user indicated in the stored section topic information and who posted the posting information of Topic _i and Topic _j . In the user information storage unit 11, the relevance calculation unit 17 posts the posting information of Topic _i within the section t, and the influence degree associated with the user who posted the posting information of Topic _j within the section t + 1. You may calculate based on. This will be specifically described below.

ここで、αは所定の値（パラメータ）を示し、ｕ_ｋ∈Ｕ_ｉ∩Ｕ_ｊはユーザ集合Ｕ_ｉとユーザ集合Ｕ_ｊとの共通部分に含まれる各ユーザｕを示し、ｎｕｍ＿ｆｏｌｌｏｗｅｒ_ｕはユーザｕのフォロワー数を示し、ｎｕｍ＿ｆｏｌｌｏｗｅｅ_ｕはユーザｕのフォロイー数を示す。αは、例えば関連トピック検出装置１に予め格納されているものとし、関連度算出部１７が当該αを適宜利用及び参照する。また、ｎｕｍ＿ｆｏｌｌｏｗｅｒ_ｕ及びｎｕｍ＿ｆｏｌｌｏｗｅｅ_ｕはユーザ情報格納部１１によって格納されたユーザ情報を関連度算出部１７が参照及び利用する。例えば、関連度算出部１７は、ユーザｕのユーザ識別情報を引数として、ユーザ情報格納部１１によって格納されたユーザ情報から当該ユーザｕの識別情報に対応付いたｎｕｍ＿ｆｏｌｌｏｗｅｒ_ｕ及びｎｕｍ＿ｆｏｌｌｏｗｅｅ_ｕを取得する。なお、ｍ＿ｆｏｌｌｏｗｅｒ_ｕ及びｎｕｍ＿ｆｏｌｌｏｗｅｅ_ｕは、値が０の場合は１とする。 When the related topic information calculation completion information is input from the related topic detection unit 15 (that is, the related topic detection unit 15 detects the Topic _{j of} the section t + 1 related to the Topic _i of the section t). ), The relevance (importance) weight (Topic _i , Topic _j ) between Topic _i and Topic _j included in the section topic information stored by the section topic information storage unit 14 is calculated by the following equation (2). .

Here, α represents a predetermined value (parameter), u _k εU _i ∩U _j represents each user u included in the common part of the user set U _i and the user set U _j, and num_follower _u represents the user u Num_followe _u indicates the number of followers of user u. For example, α is stored in advance in the related topic detection device 1, and the relevance calculation unit 17 uses and refers to α as appropriate. Further, the num_follower _u and the num_followe _u are referred to and used by the relevance calculation unit 17 with the user information stored by the user information storage unit 11. For example, the relevance calculation unit 17 uses the user identification information of the user u as an argument to obtain num_lower _u and num_followee _u associated with the identification information of the user u from the user information stored by the user information storage unit 11. In addition, m_follower _u and num_followe _u are set to 1 when the value is 0.

Note that the relevance calculation by the relevance calculation unit 17 is not limited to the calculation based on the number of followers and the number of followers. For example, the calculation may be based on the size of | U _i ∩U _j |.

The related degree calculation unit 17 updates the related topic information stored by the related topic information storage unit 16 using the calculated related degree. The table example of the related topic information illustrated in FIG. 6B illustrates a table example after the table example of the related topic information illustrated in FIG. That is, the degree of association between Topic _i and Topic _j calculated by the degree-of-association calculating unit 17 is further associated with a record indicated by a combination of the topic identification information of Topic _{i and} the topic identification information of Topic _j. Stored.

  When the calculation of the relevance level is completed, the relevance level calculation unit 17 outputs relevance level calculation completion information indicating that the relevance level calculation has been completed to the topic output unit 18.

The topic output unit 18, when the related topic detection unit 15 detects the Topic _j of the section t + 1 related to the Topic _i of the section t, based on the section topic information stored by the section topic information storage unit 14, Topic _i Out of the user set U _i that posted the posting information in the section t, the users who have not posted the posting information of Topic _{j in} the section t + 1 are extracted, and output based on Topic _j is performed for the extracted users . This will be specifically described below.

First, when the related topic information calculation completion information is input from the related topic detection unit 15, the topic output unit 18 converts the topic _i information posted to the topic _i based on the section topic information stored by the section topic information storage unit 14. From the user set U _i posted in t, the user set U _j -U _i that has not posted the posting information of Topic _{j in} the section t + 1 is extracted. For example, in the table example shown in FIG. 5, the section identification information of section t is “I001” and the topic identification information of Topic _i is “T001”, the section identification information of section t + 1 is “I002”, and the topic identification of Topic _j . When the information is “T101”, U _i is a set including users whose user identification information is “U001”, “U003”, and “U005”, and U _j is user identification information “U003” and “U004”. Therefore, the topic output unit 18 extracts a set including users whose user identification information is “U001” and “U005” as U _j −U _i .

Next, the topic output unit 18 outputs, for example, post information on Topic _j posted in the section t + 1 to the mobile terminals of each user included in the extracted user set U _j -U _i via the network. To do. For example, the mobile terminal information in which the user identification information and the mobile terminal identification information are associated with each other is stored in the related topic detection device 1 in advance, and the topic output unit 18 is connected to the mobile terminal. You may make it identify based on information. The information to be output is not limited to post information about Topic in _j, it may be any information as long as it is based on Topic in _j.

The topic output unit 18 may perform output based on Topic _j based on the relevance calculated by the relevance calculation unit 17. For example, when the relevance calculation completion information is input from the relevance calculation unit 17, the topic output unit 18 extracts the user set U _j -U _i and includes it in the user set U _j -U _i as described above. Topic _j (output target topic) is extracted for each user (output target user). Next, the topic output unit 18 includes each of the user sets U _j -U _i included in the user set U _j -U _i based on the degree of association calculated by the degree of association calculation unit 17 (related topic information stored by the related topic information storage unit 16). Output topic information that is a list in which Topic _j for each user is arranged in descending order of relevance is calculated. FIG. 7 is a diagram illustrating an example of a table of output topic information calculated by the topic output unit 18. As shown in the example of the table in FIG. 7, the output topic information includes “output target user” that is the user identification information of the output target user, “output target topic” that is the topic identification information of the output target topic, and “ “Relevance” is associated. In the table example shown in FIG. 7, the relevance level corresponds to the relevance level associated with the record whose related destination topic is the output target topic in the related topic information table example shown in FIG. The topic output unit 18 may output (or store) the calculated output topic information to the output topic information storage unit 19. Next, the topic output unit 18 performs output based on, for example, the top N output target topics with the degree of relevance for each output target user included in the calculated output topic information. Here, N is a predetermined number, for example, stored in the related topic detection device 1 in advance, and the topic output unit 18 refers to and uses the N.

Note that when the topic output unit 18 receives a determination result indicating that Topic _j is not related to Topic _i from the related topic detection unit 15, the topic output unit 18 does not perform any particular processing or displays the result as a user or the like. Output error information to be communicated to.

  The output topic information storage unit 19 stores the output topic information output from the topic output unit 18. The output topic information storage unit 19 stores, for example, the output topic information shown in the table example of FIG.

  Next, processing of the related topic detection method in the related topic detection apparatus 1 according to the present embodiment will be described using the flowchart shown in FIG.

  First, the instruction information is acquired by the instruction acquisition unit 12 (step S1, instruction information acquisition step). Next, the section topic calculation unit 13 calculates the section topic information based on the instruction information acquired in S1 and the posted information history stored in the posted information history storage unit 10 (step S2, section topic information calculation). Step), the calculated section topic information is stored by the section topic information storage unit 14. Next, the related topic detection unit 15 calculates related topic information based on the section topic information calculated in S2 (or the section topic information stored by the section topic information storage unit 14) (step S3, related Topic information calculation step), the calculated related topic information is stored by the related topic information storage unit 16. Next, the relevance level is calculated by the relevance level calculation unit 17 based on the related topic information calculated in S3 (or the related topic information stored by the related topic information storage unit 16) (step S4, relevance level). (Calculation step), the calculated relevance level is stored by the related topic information storage unit 16. Next, the topic output unit 18 related topic information calculated in S3 (or related topic information stored by the related topic information storage unit 16) and the relevance calculated in S4 (or related topic information storage unit) The output topic information is calculated based on the degree of relevance included in the related topic information stored in step 16 and output based on the calculated output topic information is performed (step S5, output topic information output step).

  The processing of the related topic detection method in the related topic detection device 1 has been described above. When section topic information is stored in advance by the section topic information storage unit 14, the processes of S1 and S2 are not essential, and the related topic detection apparatus 1 may start the process from S3. Further, the process of S4 is not essential, and when S4 is omitted, in S5, the topic output unit 18 calculates and calculates the output topic information based only on the related topic information calculated in S3. Output based on the output topic information is performed.

  Next, the effect of the related topic detection apparatus 1 configured as in the present embodiment will be described.

According to a related topic detection device 1 of the present embodiment, the relevant topic detection unit 15, interval t of Topic in _i (relation source topic) in the relevant section t + 1 of the Topic in _j (associated destination topics) is, Topic in _i and Topic in _j It is detected based on the user who posted the posting information. Thereby, for example, even when the appearance probabilities of words between Topic _i and Topic _j are different, Topic _i is more reliably determined as a topic related to Topic _i based on the user who posted the posted information of Topic _i and Topic _{j. j} can be detected. That is, a related topic can be detected more reliably.

Further, according to the related topic detection device 1 of the present embodiment, the related topic detection unit 15 includes a user who posted Topic _i posting information in the section t and a user who posted Topic _j posting information in the section t + 1. It is good also as detecting based on the similarity with. Generally, the posting information of each of two related topics is characterized by a high possibility that a user with a high degree of similarity (more similar) has posted. With this configuration, based on the similarity between the user who posted Topic _i posting information in the section t and the user who posted Topic _j posting information in the section t + 1, using the feature, Certainly, Topic _j associated with Topic _i can be detected.

Further, according to the related topic detection apparatus 1 of the present embodiment, the related topic detection unit 15 posts the posting information of Topic _i within the section t, and the user who posted the posting information of Topic _j within the section t + 1 It is good also as detecting based on the number of. Generally, the posting information of each of two related topics is characterized by the high possibility that the same user has posted. By adopting such a configuration, it is possible to more reliably use the feature based on the number of users who have posted Topic _i posting information in the section t and posted Topic _j posting information in the section t + 1. , Topic _j associated with Topic _i can be detected.

Further, according to the relevant topic detection device 1 of the present embodiment, if the Topic in _j associated with Topic in _i by Topic detection unit 15 is detected, the topic output unit 18, a post information Topic in _i in the interval t Among the posted users, users who have not posted the posting information of Topic _j within the section t + 1 are extracted, and an output based on Topic _j is performed on the extracted users. Thus, for example, of the user who made the post about Topic in _i, the user unaware of the presence of a Topic in _j related to the Topic in _i, it is possible to notice the presence of the Topic in _j. That is, it is possible to improve the convenience of the user, such as making the user aware of a topic that is related to a topic that the user is interested in and that the user is not aware of.

Also according to the relevant topic detection device 1 of the present embodiment, if the Topic in _j associated with Topic in _i by Topic detection unit 15 is detected, the relevant degree calculator 17, the degree of association between Topic in _i and Topic in _j Is calculated based on the user who posted the posting information of Topic _i and Topic _j , and the topic output unit 18 performs output based on Topic _j based on the calculated relevance. Thus, the output based on the Topic in _j associated with the Topic in _i is, Topic in _i and to be done on the basis of the relevance of the Topic in _j, for example, it can be displayed preferentially to the user a more relevance is high topics The convenience of the user can be improved.

Further, according to the related topic detection device 1 of the present embodiment, the relevance calculation unit 17 posts the posting information of Topic _{i in} the section t and the posting information of Topic _{j in} the user information storage unit 11. It is good also as calculating based on the influence degree (the number of a user's follower, the number of followers) matched with the user who contributed in the area t + 1. In general, the degree of relevance between two topics is characterized in that it is related to the degree of influence of a user who has posted on the two topics with respect to other users. With such a configuration, based on the degree of influence associated with the user who posted Topic _i posting information in the section t and posted Topic _j posting information in the section t + 1 using the feature. (Based on how much an influential posting user has transitioned between topics), the degree of association between Topic _i and Topic _j can be calculated more reliably.

  Although the present embodiment has been described using topics, the present invention is not limited to this. For example, instead of a topic, a keyword, an image, or the like may be used as information indicating the characteristics of the content of posted information.

  As described above, the related topic detection apparatus 1 of the present embodiment extracts topical structure change (inter-topic structure) using topic posting user information together with topic information learned from posts such as microblogs. To do. Next, a relevance (importance) score is added to the extracted inter-topic structure using posted user information. Finally, the posting information on the latest related topic with high importance is presented to a user who has posted or browsed on a topic. Thereby, it is possible to present related topic information having the latest and high importance. That is, real-time topic information using a microblog or the like can be presented to the user.

  FIG. 9 shows a case where a conventional method is used as a method for determining a time series change (inter-topic structure, link structure) of a topic in a situation where a topic of posting information posted by a user changes in time series, and this implementation. It is a figure which shows the comparison with the case where the method by the related topic detection apparatus 1 of a form is used.

  In the conventional method shown in FIG. 9A, the structure between topics according to time is performed by comparing the probability distribution of words in each topic. However, in reality, not only the topic changes, but it also branches into multiple topics, or multiple topics are absorbed into one topic, and a reliable inter-topic structure is only determined by comparing the probability distribution of words. Cannot judge.

  On the other hand, in the method using the related topic detection apparatus 1 of the present embodiment shown in FIG. 9B, if the topic is a related topic, even if the probability distribution of words is different, the users who are posting to the topic are similar. The topic structure is clarified by using the topic and user information posted on the topic. Furthermore, by using user information posted to a topic, it is possible to add a relevance score to the inter-topic structure, and a more accurate inter-topic structure can be determined.

  The related topic detection apparatus 1 of this embodiment can be applied in the fields of search, natural language processing, and recommendation.

  DESCRIPTION OF SYMBOLS 1 ... Related topic detection apparatus, 10 ... Posting information history storage part, 11 ... User information storage part, 12 ... Instruction acquisition part, 13 ... Section topic calculation part, 14 ... Section topic information storage part, 15 ... Related topic detection part, 16 ... related topic information storage unit, 17 ... relevance degree calculation unit, 18 ... topic output unit, 19 ... output topic information storage unit.

Claims (6)

Section topic information storage means for storing section topic information that associates a topic of posted information posted in the section with a user who posted the posted information for each section that is a predetermined time range;
The second topic of the second section related to the first topic of the first section is a user indicated in the section topic information stored by the section topic information storage means, and the first topic and the first topic Related topic detection means for detecting based on the user who posted the posting information of the two topics,
Related topic detection apparatus comprising:   The related topic detecting means is based on the similarity between the user who posted the posting information of the first topic in the first section and the user who posted the posting information of the second topic in the second section. The related topic detection apparatus according to claim 1, wherein the related topic detection apparatus detects the related topic.   The related topic detecting means detects the posting information of the first topic within the first section, and detects the posting information of the second topic based on the number of users who have posted within the second section. The related topic detection device according to claim 1.   When the second topic of the second section related to the first topic of the first section is detected by the related topic detection unit, based on the section topic information stored by the section topic information storage unit, Among the users who posted the posting information of the first topic in the first section, the users who have not posted the posting information of the second topic in the second section are extracted, and for the extracted users The related topic detection device according to claim 1, further comprising topic output means for performing output based on the second topic. When the second topic of the second section related to the first topic of the first section is detected by the related topic detection means, the degree of association between the first topic and the second topic is determined as the section. Relevance calculation means for calculating based on a user who is indicated in the section topic information stored by the topic information storage means and has posted the posting information of the first topic and the second topic,
The topic output means performs output based on a second topic based on the relevance calculated by the relevance calculation means.
The related topic detection apparatus according to claim 4. A user information storage unit that stores user information in which the user and the degree of influence of the user on other users are associated;
In the user information storage means, the relevance calculation means posts the posting information of the first topic within the first section, and the user who posted the posting information of the second topic within the second section The related topic detection apparatus according to claim 5, wherein the related topic detection apparatus calculates the influence based on the degree of influence associated with.

Images