JP2006011896A - Recommendation system, program recommendation system, and service - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To receive not messages but only index files placed for the management of a large quantity of stored messages from a group of message storage servers already operating on a network; and to read the name of a particular topic and changes in the number of messages posted to the thread of the topic in a given period of time and detect any abnormal increase in the number of messages posted to a program or the like estimated from the name of the topic, thus recommending the program. <P>SOLUTION: Only the index files that are located in a group of servers processing messages sent over a network and that include not the contents of messages posted but at least the names of threads or the number of representative messages and the number of posts are read, whereby any abnormal increase in the number of messages posted to a particular thread or topic is detected and the name of a currently popular program is estimated from the name of the thread or topic and recommended. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention efficiently provides the user with information that is popular but playful information, although the information is rarely interesting for the user.

  With the development of the Internet, users are unable to follow the flood of various information, and various information retrieval systems have been developed. In addition, the provision of search engines and the like for searching information widely and shallowly by inputting natural sentences has become a reality.

  From the viewpoint of programs that can be viewed, digitalization of television broadcasts and radio broadcasts / communications has progressed. Furthermore, broadcasts / communications via networks such as the Internet have emerged, and only terrestrial analog broadcasts are performed. The number of channels that can be viewed has increased significantly compared to the times. Therefore, each broadcasting station produces programs of various genres, and even programs that meet the needs of a small number of users are produced and broadcast. As recording media become larger, programs continue to be recorded, and users want guidance on which programs to watch during their valuable time.

  On the other hand, Patent Document 1 discloses a first conventional technique using “viewing rate” as a criterion for selecting a program. According to this technique, the user can know the audience rating of a program, that is, how many users are viewing the program, and can select a program based on this information.

  On the other hand, Patent Document 2 discloses a second prior art in which a keyword is given to a program and the degree of matching with a keyword registered in advance by a user is used as a material for determining program selection. According to this technique, the user can select a program according to his “interest”.

  Patent Document 3 discloses a large amount of information (messages) about a program that is transmitted / provided by an unspecified number of users such as viewers and listeners (hereinafter referred to as viewers) on a network such as the Internet. A third prior art is disclosed in which the program interest level of a general user is digitized by receiving, storing, and analyzing, and thereafter a program interest level presentation service is performed from a server. According to this technology, the user can receive services such as information that is popular but playful information, although the information is rarely interesting for the user.

  In Non-Patent Document 1, an infinite state transition machine is used as a more detailed analysis method than Patent Document 3, and an abnormal increase in keywords is specified from a large number of messages flowing in time series. As a result, more specific numerical values can be obtained, including not only the program interest level of general users but also the interest level of scenes in the program.

In FIG. 6, typical documents are illustrated.
JP-A-9-261609 JP-A-11-187328 JP 2002-279026 A "Bursty and Hierarchical Structure in Streams" by Jon Kleinberg, Proc. 8th ACM SIGKDD Intl. Conf. On Knowledge Discovery and Data Mining, 2002. [Search April 2004] Internet <a href = "http: // www. cs.cornell.edu/home/kleinber/bhs.pdf ">

  However, the “viewing rate” employed in the first prior art is not the best program selection criterion. In other words, “viewing rate” is only an indicator of the number of users who are watching a program, and in many cases, how much “interest” each user is watching, that is, the user ’s interest in the program. This is because it does not accurately indicate “degree of interest”. In other words, even if the majority of users simply turn on the TV, the majority of users are watching the program with a high degree of interest in what is called a program that seriously considers information from the program. However, if the number of viewers of both programs is the same, the “viewing rate” will be the same value. In particular, when the number of broadcasting / communication channels increases, the “viewing rate” of each program is expected to be extremely low compared to the “viewing rate” of the currently broadcast / communication program. There is a possibility that there is no big difference in "viewing rate". Also, when hard disk recording technology has become widespread, the current audience rating, which is calculated by counting only the number of users who are viewing in real time, also provides accurate information on how many viewers there are. There is a possibility that it cannot be expressed.

  On the other hand, in the second conventional technique focusing on the “interest” of the user, a keyword is added to the program so that a program that matches the keyword registered in advance by the user can be selected. However, user interest usually changes from moment to moment. Therefore, in order to select an appropriate program, the user always needs to grasp what is his “interest”, search for a keyword matching this “interest”, and register it. . Also, users do not always grasp their own “interests” accurately and select and watch programs that match these “interests”, but in many cases, they are programs that others are interested in. If there is, the user selects and watches the program based on an empirical rule that the person should be interested. Therefore, it can be said that the “viewing rate” employed in the first prior art has been developed to satisfy such user needs.

  On the other hand, on a network such as the Internet, various information is transmitted and transmitted using, for example, a Web page. For example, so-called chat technology such as electronic bulletin board (BBS (Bulletin Board System)) technology and IRC (Internet Relay Chat (RFC1495, 2812)) technology, e-mail technology, mailing list technology, etc. are rapidly spreading in recent years. It is also possible to communicate directly. Some users use these technologies to communicate and transmit information such as program content introductions and reviews. Since this information is not information by a program sender (broadcast station or the like) but information by a viewer, it can be said to be objective information and extremely useful as a guideline for selecting a program. In particular, it can be said that a user who transmits information about a program using such means has a very high “degree of interest” for the program. Therefore, the number of information transmissions for the program, as well as the content of this information, is considered to be a very useful index indicating the “interest level” of other users' programs.

  In the third prior art described above, a large amount of messages for a program transmitted / provided by a viewer on a network such as the Internet is received, stored, analyzed, and the “interest” of the viewer's program is quantified to show the program interest level. The server group to be presented is operated, and the service is provided for the user to select the program. However, the network load is too large to receive a large amount of messages flowing on the network, and in order to analyze a large amount of messages, it is necessary to operate a server group, which is not something that individual users can do freely. In addition, it is difficult to recommend a program that can be viewed in real time because it is necessary to move a lexical analyzer to search for related messages.

  In the fourth prior art, a technique for specifically detecting an abnormal increase in keywords from a large number of messages distributed on the network using a state transition machine is described. However, only a more precise technique is used, and the analysis time is longer than that of the third conventional technique, and real-time characteristics cannot be obtained.

  The present invention solves the above-mentioned conventional problem, and receives only an index file placed for managing a large amount of accumulated messages, not a message itself, from a group of message accumulation servers already operated on a network. An object of the present invention is to read a change in the number of posted messages per hour for a specific topic name and a thread formed by the name, detect an abnormal increase in the number of posted messages for a program estimated from the topic name, and recommend the program.

  In order to solve the above-described conventional problem, the recommendation method of the present invention includes at least a thread name or a representative message subject and the number of posts in a server group that processes a message transmitted on a network and does not include posted message content. By detecting only an increase in the number of messages posted to a specific thread or topic by reading only the included index file, the name of the currently popular one is estimated from the name and recommended.

  With this configuration, what is popular in real time can be recommended to the user.

  According to the recommendation method of the present invention, what is currently popular can be recommended in real time.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(Embodiment 1)
FIG. 1 is a configuration diagram of a recommendation method according to Embodiment 1 of the present invention. In FIG. 1, reference numerals 2 to 14 denote components of the present invention, and reference numeral 11 denotes a system for managing a group of messages posted to a network outside such as an electronic bulletin board.

  Details of the electronic bulletin board system 11 outside the present invention will be described with reference to FIGS.

  4 shows a physical configuration such as an electronic bulletin board, 41 is a data stream input unit for receiving message groups posted on the network, 42 is a means for storing a large number of received message groups, and 43 is a large number of message groups. A server group for management, 44 is a small index file group created by the server 43 for management, and 45 is a web server or the like that responds to external requests and transmits various information currently stored in an electronic bulletin board or the like , 46 represents the entire system constructed based on the recommendation method of the present invention. Before referring to the contents of the message group, the management system 43 first refers only to the index file, starts with managing the number of messages included in each thread, and then logically includes multiple stages or trees such as message titles. The management of the structure is made possible and will be described with reference to FIG.

  FIG. 5 shows a logical configuration of a system such as an electronic bulletin board. A server 51 is usually virtualized at the entrance of the entire system, and its physical entity is usually distributed among a plurality of server groups. 52 is also called a “plate”, for example, a thread group generic name divided by field. In a huge electronic bulletin board system, there is a server that is a physical entity at least for each board. In addition, there is a thread group 54 for each “board”. For example, in the “actual thread board”, there is a drama “pride” thread, and a large number of messages are posted to the thread especially during the program broadcast. In particular, in a scene where a popular actor appears, a short message saying “Kim Tak Kita!” From a single unspecified number of people is posted in real time while the program is being broadcast, and the more popular programs the greater the number of posts. There are hundreds of messages posted per minute in the scenes that attract attention. On the other hand, in order to manage a large number of messages, the bulletin board system does not include message contents called index files, but includes the board name, the list of thread names included in it, the total number of messages posted to each thread, and other options. In general, a small management file including only the thread id (unique number) and the like exists for each normal board, and the physical entity thereof is the index file group of the 44.

  Returning to FIG. 1, the recommendation method in the first embodiment of the present invention will be described. 12 is an index request / input unit for requesting 11 to transmit the index file and receiving the index file from 11, 13 is requesting 12 to receive the index file at an appropriate time interval, and 12 is received It is a message number abnormality increase detection unit that limits a specific thread (topic) in which the number of posted messages is abnormally increased by reading an index file and calculating the number of message posts per hour to each thread (topic). For example, by receiving the index file every five minutes and searching for a specific thread in which the number of messages posted on the network has increased rapidly in the five minutes. 14 is a recommendation unit that recommends the name of the thread detected in 13 or the message subject representing the thread as a topic that is currently in fashion, or whether to actually recommend it to the user in view of the user profile or the like. is there. In addition, as a user profile application, it is also possible to check a field in which the user is interested first, and restrict the index file of the thread group belonging to the field to be accepted from 11.

  By using such a method, it is not necessary to judge the contents of a large number of messages posted on the network, and only a small index file for managing it is processed, and the index file is analyzed and analyzed for a specific topic. By presenting an abnormal increase in the number of posted messages and recommending the topic as one that is currently in the public interest, it presents public interests or trendy items that do not cause the user to think without overloading the network. Make it possible. Of course, it is also possible to make a recommendation after selecting a thing close to the user's own interest in consideration of the user profile and the like. For example, it can be used only for a program in which actors such as SMAP members who are closely related to the actor Takuya Kimura are recommended.

  In the present embodiment, 11 is provided as an external database system, but this may be incorporated into the present invention. For example, a server function such as an electronic bulletin board may be placed inside the present invention as 11, and a message posting of a like person may be invited.

  In addition, there are other boards (thread groups) that handle program information in the electronic bulletin board. For example, there is a “TV board”, so not only the threads in the “realistic thread board” but also related threads in multiple boards. It is also preferable to consider the total number of messages posted.

(Embodiment 2)
FIG. 2 is a configuration diagram of the program recommendation method according to the second embodiment of the present invention. In FIG. 2, 12-14 and 22-25 are the components of the present invention. In addition, the same code | symbol is used about the same component as FIG. 1, and description is abbreviate | omitted.

  In FIG. 2, 21 is an external transmission source system for electronic data such as electronic program information (EPG) that can be easily obtained separately, and 22 is a program guide for receiving data transmitted by 21 in the present invention. A request / input unit 23 is an on-air program extracting unit that selects the currently aired program from the electronic program information received at 22. 24 is a collation unit that collates the currently broadcast program information extracted from 23 with the current popular topic detected from 13 and identifies a program that is currently being broadcast and attracted public attention. is there. This is the name of the thread showing an abnormal increase in the “actual thread board” obtained from the message number abnormal increase detection unit 13, but is usually the program name itself or its abbreviation, and is obtained from 23 It is easy to collate with the program information that is currently airing. For example, in the case of the Korean drama “Winter Sonata”, the thread name is usually simply abbreviated as “Winter Sonata”, and the fuzzy matching function is sufficient. Furthermore, when using electronic program information, etc., not only the program name but also the genre to which the program belongs can be obtained. Therefore, the recommendation unit 14 recommends the user profile after limiting it to the genre preferred by the user. Easy. Reference numeral 25 denotes a user presenting unit obtained at 14 that can present not only a program that is currently attracting interest in the world but also a specific scene in which an abnormal increase in the number of posted messages is detected. In addition, an indicator may be added to the screen to show a specific change in the number of posted messages in a scene.

  In the present embodiment, the example in which the collation by the collation unit 14 of the detected thread name is performed using the EPG data has been described, but the collation data is not limited to this. For example, a search query may be sent to an external search service and the result may be used. Alternatively, an associative dictionary defining relationships between words may be provided inside or outside, and other keywords associated with the detected thread names may be recommended.

(Embodiment 3)
FIG. 3 is a configuration diagram of a program recommendation method and service according to the third embodiment of the present invention. In FIG. 3, 12-14, 22-24, and 33 are constituent elements of the recommendation service in the present invention, and the addition of 25 and 32 constitutes a recommendation method in the present invention. At this time, the communication path connecting 25 and 33 may be a network such as the Internet outside the present invention or may be included in the present invention. In FIG. 3, the same components as those in FIGS. 1 and 2 are denoted by the same reference numerals, and the description thereof is omitted.

  In FIG. 3, reference numeral 31 denotes a system outside the present invention that distributes a program or the like that the user finally wants to view, such as a broadcasting station. Reference numeral 32 denotes means for actually storing programs distributed by 31. It is also possible to store at least all the latest programs like a FIFO memory. When the storage means 32 is present on the user side, the user does not need to view the program in real time, so-called recording viewing is possible. Therefore, the following changes can be made from the first and second embodiments. First, in the specific scene search in which an abnormal increase in the number of posted messages is detected especially during the program broadcast, the recommendation unit 14 is, for example, when 12 requests an index only once every five minutes, It should be possible to present it to the user or record and view it retroactively up to less than 5 minutes. Next, since the recommendation system itself does not necessarily have to be on the user side, only information on what the recommendation unit 14 recommends, for example, a recommended program by directly sending to 33 electronic news, e-mail, or a user-side system port, for example, a recommended program It becomes possible to introduce only the name as a delivery service to the user. In addition, it is possible to make it a recommended program information distribution / uM service only for registered users by combining it with the user management system. The user profile is registered in advance, and based on this, the recommendation unit 14 provides each user with a user profile. It is also possible to filter and recommend. Conversely, IRC or peer-to-peer (P2P) may be used to allow all participants to share information. The user presentation unit 25 in FIG. 3 multiplies the program name information recommended by the recommendation unit 14 and the program name information included in the program stored in the program storage unit 32 and recommends only the stored program. It is adjusted so that it can. Alternatively, the storage unit 32 itself may cooperate with the recommendation unit 14 from the beginning, and the algorithm may be changed so that at least only the recommended program is not automatically deleted but stored, instead of storing all the programs.

  The recommendation method according to the present invention has a capability of analyzing a posted message group flowing on a network, and is useful as a method for recommending a general fashion or the like. It can also be used for recommending programs.

The figure which shows the recommendation apparatus of embodiment of this invention The figure which shows the recommendation apparatus of embodiment of this invention The figure which shows the recommendation apparatus of embodiment of this invention Diagram of electronic bulletin board system of the present invention A diagram showing the logical structure of an electronic bulletin board, etc. The figure which shows the conventional recommendation device

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Electronic bulletin board system 12 Index file request | requirement and input part 13 Posted message number abnormal increase detection part 14 Recommendation part 21 Electronic program guide information etc. system 22 Electronic program guide request | requirement and input part 23 Broadcasting program etc. extraction part 24 Collation part 25 User presentation unit 31 Broadcast station etc. system 32 Program etc. storage means 33 Electronic news, email, IRC etc. distribution system 41 Post message etc. data stream input part 42 Post message storage means 43 BBS etc. management server group 44 Index file 45 Web server etc. System 46 Recommendation system 51 Virtual server 52 Board 53 Index 54 Thread group

Claims (14)

By reading an index file that contains at least the thread name or representative message subject, and the number of messages posted to the thread, in the servers that process messages sent over the network,
Detects an increase in the number of messages posted to a specific thread,
A recommendation method characterized by recommending the thread name or the representative message subject, or the one estimated from the name. The recommendation method according to claim 1,
A recommendation method using a sum of related threads as the number of posted messages. The recommendation method according to claim 1,
A recommendation method characterized by limiting search target threads from a user profile. The recommendation method according to claim 1,
A program recommendation method characterized by recommending a suitable program as a result of collating thread information having a large number of posts per hour with program information currently being broadcast. 5. The program recommendation method according to claim 4, wherein a moment at which the posting number increase speed is large or a time just before that is recommended as an attention scene in the program. The program recommendation method according to claim 4, wherein EPG (electronic program information) is used as program information. 5. The program recommendation method according to claim 4, wherein the recommended program is retained from the storage means without being automatically deleted. By reading an index file that contains at least the thread name or representative message subject, and the number of messages posted to the thread, in the servers that process messages sent over the network,
Detects an increase in the number of messages posted to a specific thread,
Manage the increase over time,
A program recommendation service characterized in that, as a result of collating thread information having a large number of postings per hour with currently broadcast program information, a compatible program is distributed as recommended program information. The program recommendation service according to claim 8, wherein electronic mail is used as a distribution system. The program recommendation service according to claim 8, wherein peer-to-peer is used as a distribution system. The program recommendation service according to claim 8, wherein electronic news is used as a distribution system. A recommendation method according to claim 1, wherein an electronic bulletin board system is used as a server. The recommendation method according to claim 1, wherein a mailing list system is used as a server. The recommendation method according to claim 1, wherein an Internet relay chat system is used as a server.

Images