JP2018067041A - Extraction apparatus and computer program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an extraction apparatus and a computer program capable of appropriately extracting a person conforming to a desired person image.SOLUTION: An extraction apparatus includes a reception unit that receives a text content corresponding to a desired person image, a storage unit that stores a transmission information database for recording a plurality of transmission information transmitted by a plurality of transmitters respectively in association with identification information identifying each of the transmission information and a transmitter identification information database for recording a transmitter identification information identifying each of the transmitter of the transmission information, and a calculation unit that calculates a similarity between a group of the transmission information transmitted by the transmitter and the text content received by the reception unit using the transmission information database and the transmitter identification information database for each the transmitter, and an extraction unit that extracts the transmitter identification information of the transmitter whose similarity calculated by the calculation unit is equal to or more than a predetermined degree from the transmitter identification information database.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an extraction apparatus that extracts a desired caller based on transmission information on a network community and a computer program that causes a computer to operate as the extraction apparatus.

  As the quality and function of products and services mature, the differentiation of products and services depends on a wide variety of consumer preferences such as design, texture, and concept image. In developing products and services that meet a wide range of consumer preferences, it is said that it is effective to incorporate the opinions of consumers so that they can be supported by as many consumers as possible. For this reason, in addition to referring to questionnaires, reviews, comments, etc. from consumers, efforts such as recruiting participation in product development have been made.

  In Patent Document 1, ideas exchanged by chat between members registered in advance with respect to the product (sometimes with the product development operator), and voting for image illustrations, etc. are also made via the network. A system to be realized is disclosed.

  Patent Document 2 discloses a system (monitor house) that provides a place for product development to applicants who wish to commercialize monitor products. In the monitor house proposed in Patent Literature 2, a place for exchanging questionnaires with applicants, members registered in advance and supporters for commercialization of monitor products, or requests for advertisement after commercialization A place is provided. At this time, it is disclosed that it is possible to narrow down the members or supporters targeted by the questionnaire based on conditions such as sex and age.

Japanese Patent Laid-Open No. 2003-091628 JP 2001-357194 A

  When performing idea exchange, questionnaire, voting, etc. for a person who is a consumer as shown in Patent Document 1, as a result of the idea exchange, the result of the vote is what kind of member the member who shared the idea Depending on whether or not the result is different. Therefore, as disclosed in Patent Document 2, a business operator who develops a product widely narrows down the target person of the questionnaire by attribute according to what consumer is the target of the product. .

  However, when narrowing down the consumers (customers) that should be the target of products and services to be developed based on attribute information such as keywords representing gender, age, region, and preference, it is not possible to extract an appropriate person due to some problems. there is a possibility. First, in the development of diversifying products and services, it is necessary to classify the attribute information into a huge number in order to express the target consumer image with the attribute information. Therefore, it becomes very complicated for a general consumer to register the attribute information by self-report, and accuracy is lost. Furthermore, it is difficult to measure the strength and degree of each interest / interest in the attribute information, and it is difficult to distinguish between a person who is lightly interested and a person who is strongly interested / interested. Therefore, in the extraction based on the attribute information, there is a possibility that a person who matches the person image to be targeted, that is, a person other than a person who is really interested or interested in the target product may be extracted. In addition, it is difficult to reflect temporal elements in the attribute information. For example, if a keyword in a field that has already lost interest remains registered as the attribute information of the person, there is a possibility that a person who is no longer interested or interested at that time may be extracted.

  The present invention has been made in view of such circumstances, and an object thereof is to provide an extraction apparatus and a computer program that can appropriately extract a person in accordance with a desired person image.

  The extraction apparatus according to the present disclosure associates a reception unit that receives text content corresponding to a desired person image, and a plurality of pieces of transmission information that are respectively transmitted from a plurality of senders, with identification information that identifies each piece of transmission information. A recorded transmission information database, a storage unit for storing a sender identification information database in which sender identification information for identifying each sender of the transmission information is recorded, and transmission using the transmission information database and the sender identification information database A calculation unit that calculates the similarity between the transmission information group transmitted by the caller and the text content received by the reception unit, and the similarity calculated by the calculation unit is equal to or greater than a predetermined level. An extraction unit that extracts caller identification information of the caller from the caller identification information database.

  In the extraction device according to the present disclosure, the extraction unit sorts the caller identification information in descending order of the corresponding similarity.

  The extraction device according to the present disclosure includes, for each caller identification information extracted by the extraction unit, information related to the calculation of the similarity and display information for displaying a plurality of transmission information contributing to the similarity. A creation unit is further provided.

  In the extraction device according to the present disclosure, a transmission time of transmission information is recorded in the transmission information database in association with the transmission information, and the calculation unit transmits a transmission transmitted by the caller for each caller. Extracting the information group, narrowing down the transmission information group transmitted in a predetermined length from the extracted transmission information group, and the text content received by the reception unit using the narrowed transmission information group The similarity is calculated.

  In the extraction device according to the present disclosure, a transmission time of transmission information is recorded in the transmission information database in association with the transmission information, and the calculation unit is configured to transmit each transmission information extracted from the transmission information database. The transmission time is assigned a weighting coefficient that becomes a higher numerical value in the order from the calculation time of the similarity, the assigned weighting coefficient is multiplied by the number of appearances of words included in the transmission information, and the transmission is performed for each sender. The similarity with the text content received by the reception unit is calculated using the words included in the transmission information group transmitted by the person and the number of appearances of the words.

  In the extraction device according to the present disclosure, the reception unit receives a plurality of text contents together, and the calculation unit calculates a similarity for each of the plurality of text contents, or is derived from the plurality of text contents. The similarity is calculated using the information indicating.

  In the extraction device according to the present disclosure, the calculation unit is extracted from a related dictionary in which words included in the transmission information or the text content received by the reception unit and related words related to the words are recorded. Similarity is calculated using related terms.

  The computer program according to the present disclosure includes a transmission information database in which a plurality of transmission information respectively transmitted from a plurality of senders is recorded in association with identification information for identifying each transmission information, and each sender of the transmission information. A computer program for causing a computer capable of reading and writing to a sender identification information database recording sender identification information to be extracted to extract the sender identification information, the text content corresponding to a desired person image on the computer Calculating the similarity between the transmission information group transmitted by the sender and the received text content for each sender using the transmission information database and the sender identification information database. The caller identification information of the caller whose similarity is equal to or higher than a predetermined level is sent. To execute the steps of extracting from the identification information database.

  In the extraction device of the present disclosure, a person in accordance with a desired person image that matches the text content is appropriately extracted by the process of extracting the sender identification information of the sender who transmits the transmission information similar to the text content. It becomes possible to do.

It is a schematic diagram which shows the outline | summary of the information processing system which concerns on this Embodiment. It is a block diagram which shows the internal structure of each apparatus which comprises an information processing system. It is explanatory drawing which shows the example of the content of user information DB. It is explanatory drawing which shows the example of the content of transmission information DB. It is a flowchart which shows an example of the procedure of the extraction process by a server apparatus. It is explanatory drawing which shows the example of the content of transmission information. It is explanatory drawing which shows the example of the content of a text content. It is a flowchart which shows an example of the procedure of the creation process in which a control part produces the display information which displays an extraction result. It is explanatory drawing which shows the example of a screen displayed on the display part of a client apparatus based on the display information produced by the control part. It is a flowchart which shows another example of the procedure of the extraction process by a server apparatus. It is a flowchart which shows another example of the procedure of the extraction process by a server apparatus. It is a flowchart which shows another example of the procedure of the extraction process by a server apparatus.

  The present invention will be specifically described with reference to the drawings showing the embodiments thereof.

  FIG. 1 is a schematic diagram showing an outline of the information processing system according to the present embodiment. The information processing system includes a server device 1 and client devices 2 and 3 that can be connected to the server device 1 for communication. The server apparatus 1 has a server function that provides a network community 100 that allows a plurality of general users to post text contents such as chat, writing on a bulletin board, comments, and reviews via the client apparatus 2. Then, the server device 1 matches a person image desired by a specific user who is a product developer based on transmission information (speaking, writing, voice, blog article, etc.) V from a plurality of general users on the network community 100. It functions as an extraction device that performs processing for extracting a general user.

  In the example shown in FIG. 1, user D who is a product developer wants to obtain the cooperation of a general user who is the target of a product in developing a unique product using a specific food that has become a boom. Think. For example, there are users A, B, and C as general users who speak and write in a group relating to gourmet on the network community 100. It is assumed that the user A is a person who has been very fond of the specific food for a long time, but prefers to go to the store where the local cook provides food or travel to the home. It is assumed that the user B is a person who frequently cooks at home, while being widely interested in what has become a boom including the specific ingredients. The user C is also a person who likes the specific ingredients but often enjoys eating out. On the other hand, the user D who is a product developer is interested in foods that are booming and desires to extract a person such as the user B who is familiar with cooking at home.

  In such a case, for example, if a user is extracted using the name of a target specific food as attribute information, even if user A and user C can be extracted, the user has a special strong interest in the specific food. There is a possibility that no user B is extracted. In the information processing system according to the present embodiment, the text content S that the user who wants to target is likely to like is received from the user D via the client device 3 without using the attribute information, and extracted using the text content S. To do. The text content S may be not only articles in newspapers, magazines or electronic media, but also text or audio or moving images. In the example shown in FIG. 1, for example, the text content S is an article about the appeal of growing booming ingredients at home, and cooking and eating them. The server device 1 appropriately extracts the user B who seems to be close to the desired target person image of the user D based on the text content S and the transmission information V from the users A, B, and C in the network community 100. can do. Furthermore, when the server apparatus 1 provides an opportunity for communication between the user D and the user B, the user D can obtain the cooperation of the user B in product development.

  In order to realize such an appropriate person extraction, the server apparatus 1 (extraction apparatus) in the information processing system has the following configuration, for example. FIG. 2 is a block diagram showing an internal configuration of each device constituting the information processing system. The server device 1 is communicatively connected to the client devices 2 and 3 via the communication medium 4.

  The communication medium 4 includes a LAN 41, a public network 42 such as the Internet, an access point (AP) 43 to the public network 42, a carrier network 44 provided by a communication carrier, and a base station 45 for accessing the carrier network 44. The communication medium 4 connects the server apparatus 1 and the client apparatuses 2 and 3 through a wired or wireless communication connection.

  The server device 1 uses a server computer and includes a control unit 10, a storage unit 11, a temporary storage unit 12, and a communication unit 13. The control unit 10 uses a CPU (Central Processing Unit), a clock, and the like, executes each process based on the server program 11P and the extraction program 12P stored in the storage unit 11, and sets the general-purpose server computer as a community providing server and an extraction device To function as. The temporary storage unit 12 temporarily stores information generated by the processing of the control unit 10 using a volatile memory such as a DRAM (Dynamic Random Access Memory).

  The storage unit 11 stores data referred to by the control unit 10 in addition to the server program 11P and the extraction program 12P using a hard disk. The storage unit 11 stores user information of the network community 100 created by the control unit 10 as a user information (sender identification information) DB 111, and stores transmission information V on the network community 100 as a transmission information DB 112. Note that the location of the user information DB 111 and the transmission information DB 112 is not limited as long as the control unit 10 can read and write information, and the user information DB 111 and the transmission information DB 112 may be configured to be stored in a storage device outside the server device 1.

  The server program 11P is a server program for causing the server computer to function as a chat server, a bulletin board server, or a content posting server. The extraction program 12P is a program for causing the control unit 10 to execute processing that will be described later for extracting user identification information.

  The communication unit 13 is a network card connected to the LAN 41 included in the communication medium 4. The control unit 10 can communicate with the communication unit 13 via the communication medium 4. For example, the control unit 10 can communicate with the client apparatuses 2 and 3 via the public network 42 via the router.

  The client device 2 and the client device 3 use a smartphone, a tablet terminal, a desktop type, or a laptop type PC. The client devices 2 and 3 can be applied to any information terminal having a communication function via the communication medium 4 such as an information terminal called a book reader, a game machine, or a PDA. The client device 2 and the client device 3 basically have the same components, and include a control unit 20 (30), a storage unit 21 (31), a temporary storage unit 22 (32), a display unit 23 (33), and an operation unit. 24 (34), voice input unit 25 (35), and communication unit 26 (36). The difference in code corresponds to the difference in interface according to the difference in the type of user to be used. The common configuration will be described below in the client device 2, and different interfaces will be described later.

  The control unit 20 includes a CPU, a clock, and the like, executes each process based on the client program 2P stored in the storage unit 21, and causes the general-purpose computer to function as the client device 2. The temporary storage unit 22 temporarily stores information generated by the processing of the control unit 20 using a volatile memory such as a DRAM.

  The storage unit 21 uses a non-volatile memory such as a hard disk or a flash memory. In addition to the client program 2P, the storage unit 21 stores general-purpose programs for clients such as a Web browser program, and further stores various data read and written by the control unit 20.

  The client program 2P is a program that causes the control unit 20 of the client device 2 to execute each process, as will be described later. The client program 2P may be a client program recorded on a recording medium (not shown) that is read by the reading unit or acquired and recorded via the communication unit 26 (none of which is shown).

  The display unit 23 uses a touch panel built-in display. The control unit 20 displays various operation screens including images such as text and icons on the display unit 23 based on the client program 2P. The display unit 23 may be a display that does not have a built-in touch panel.

  The operation unit 24 uses a touch panel built in the display of the display unit 23 and a button group provided on the housing of the client device 2. When the client device 2 is a PC, the operation unit 24 includes a user interface such as a keyboard and a mouse. The operation unit 24 notifies the control unit 20 of operation information by the user.

  The voice input unit 25 is a microphone. The control unit 20 inputs voice from the voice input unit 25. The control unit 20 can convert the voice input by the voice input unit 25 into text by voice recognition.

  The communication unit 26 includes a network card that can be connected to a LAN cable and is connected to the public network 42, or a wireless communication module based on a communication standard connected to the base station 45 and a wireless communication module corresponding to connection to the AP 43. including. The control unit 20 can communicate with the server device 1 via the communication medium 4 by the communication unit 26.

  Among the client devices 2 and 3, the client device 2 used by a general user provides a connection interface (GUI) to the bulletin board and chat room on the network community 100 provided from the server device 1 by the client program 2P. The The network community 100 is, for example, a questionnaire community derived from an information providing site that is widely disclosed on the public network 41 for various products provided by the user D, and only the terminal device 2 of the user having login information is provided. Communication connection is possible. The client device 3 used by the business operator such as the user D is provided with an interface for presenting the total results of various information (questionnaire, voting, etc.) in the questionnaire community. Further, the client device 3 includes an interface for realizing information exchange as will be described later with the operator of the server device 1 that provides the network community 100.

  FIG. 3 is an explanatory diagram showing an example of the contents of the user information DB 111. In the user information DB 111, user identification information (user ID) for mutually identifying users in the network community 100 is stored as one record in association with a user name (display name) and login information (password, etc.). That is, the user information DB 111 corresponds to a database of information for identifying the sender of the transmission information V. The login information is login information to the network community 100. In the example shown in the explanatory diagram of FIG. 3, user identification information that can be distinguished and classified into general users and business users is assigned to the users. In the example of FIG. 3, user identification information that is a 5-digit serial number starting with “0 (zero)” is given to a general user, and a user identification that is a 5-digit serial number in the 80,000 range is assigned to a business user. Information is given. The users A, B, C, and F, which are general users, further store contact information such as e-mail addresses and addresses, as well as login histories, as in the example shown in the explanatory diagram of FIG. May be. Further, as in the example shown in the explanatory diagram of FIG. 3, contact information such as an e-mail address and an address, a login history, and the like may be stored for the users D and E who are business users. Furthermore, in one record, user attribute information (gender, age (date of birth), and preference questionnaire results) may be stored together.

  The transmission information V on the network community 100, for example, the message on chat, the writing on the bulletin board, the product review, and the comment are stored in the transmission information DB 112 each time. FIG. 4 is an explanatory diagram showing an example of the contents of the transmission information DB 112. In the transmission information DB 112, text data indicating the contents of the transmission information V, transmission information identification information for mutually identifying each transmission information V, and user identification information for identifying each sender are stored in association with each other. As shown in FIG. 4, the transmission information DB 112 may store the transmission time (writing, posting time) in association with each other.

  Next, a process of extracting user identification information of a person who matches a desired person image by a request in the server apparatus 1 will be described. FIG. 5 is a flowchart illustrating an example of the procedure of extraction processing by the server device 1. The processing procedure shown in the flowchart of FIG. 5 is started in the following case, for example. First, a user D who is a business user logs in to the network community 100 from the client device 3. The top page after log-in for the business operator displayed on the display unit 33 of the client device 3 includes an interface display for accepting an “extraction request”, and is started when this “extraction request” is selected. . Alternatively, when an e-mail having a predetermined format is transmitted together with the text content S to the administrator of the network community 100, the server apparatus 1 may start the following processing upon receiving this e-mail.

  An interface for accepting an “extraction request” is displayed on the display unit 33 of the client device 3 (step S301). The interface is, for example, a Web page, and includes a selection icon for the text content S to be uploaded. In addition, a link (icon) to a detailed setting page as described later may be included.

  The control unit 30 accepts selection of the text content S through the displayed interface (step S302). The control unit 30 uploads the selected text content S to the server device 1 from the communication unit 36 together with a general user extraction request (step S303). The upload may be data of the text content S itself or link information to the text content S. In step S303, the processing in the client device 3 is temporarily terminated.

  The control unit 10 of the server device 1 receives the text content S through the communication unit 13 (step S101). In step S101, when the data of the text content S transmitted from the client device 3 is not the document data of the article itself but link information to the Web page, the control unit 10 downloads the document data from the link destination. . Further, the data of the text content S transmitted from the client device 3 may be audio or moving images, and the control unit 10 may perform the process of converting the text here in advance.

  The control unit 10 performs morphological analysis on the received text content S (step S102), derives information indicating the characteristics of the text content S (step S103), and stores the information in the storage unit 11 or the temporary storage unit 12. (Step S104). The information indicating the characteristics in step S103 is, for example, frequent nouns and the number of appearances thereof.

  Next, the control unit 10 acquires a plurality of pieces of user identification information corresponding to general users from the user information DB 111 of the storage unit 11 (step S105), and selects one by one according to a predetermined condition (step S106). The predetermined conditions are, for example, conditions such as all user identification information recorded in the user information DB 111 or user identification information recorded in association with a predetermined group (bulletin board, chat room). In addition, the user identification of only users who have the number of outgoing information over a predetermined period over the entire recorded period or the most recent predetermined period (for example, 1 month, 3 months, etc.) It may be a condition such as information. Furthermore, it may be a condition of user identification information of a user who is transmitting transmission information including a predetermined word or a user who is transmitting transmission information including a predetermined word in a predetermined period.

  The control unit 10 extracts the transmission information V group having the user identified by the selected single user identification information as the sender from the transmission information DB 112 (step S107), and performs morphological analysis on all the extracted transmission information. (Step S108). The control unit 10 derives information indicating the characteristics of the transmission information transmitted from the user of the selected user identification information based on the transmission information V group extracted based on the result of the morphological analysis (step S109), and the storage unit 11 or the temporary storage unit 12 (step S110). The information indicating the feature in step S109 can be compared with the feature information derived in step S103 in the subsequent step S111. For example, it is a frequent noun and the number of appearances.

  The control unit 10 compares the information indicating the characteristics of the text content S with the information indicating the characteristics of the extracted transmission information, calculates the similarity, and stores the similarity calculated in association with the user identification information in the storage unit 11 or It memorize | stores in the temporary memory part 12 (step S111). For example, the degree of similarity is obtained by calculating a cosine similarity between vectors obtained by using nouns having a predetermined number of appearances or more as components, or by calculating a Euclidean distance. The calculation of the degree of similarity is not limited to this, and a known invention for determining the degree of similarity and the degree of association between words in natural language processing may be used.

  Next, the control unit 10 determines whether or not all the user identification information of the user that matches the predetermined condition has been selected (step S112). When it is determined that there is unselected user identification information (S112: NO), the control unit 10 returns the process to step S106, and selects the next user identification information according to a predetermined condition (S106).

  If it is determined in step S112 that all have been selected (S112: YES), the control unit 10 sorts the user identification information associated with the similarities in descending order of similarity (step S113). The control unit 10 transmits the extraction result obtained by narrowing down the user identification information sorted in descending order of similarity to, for example, the top 30 items from the communication unit 13 to the client device 3 (step S114), and ends the extraction process. . The transmission of the extraction result in step S114 is transmitted after being created as a web page (link information) that can be referred to from the top page after the log-in of the business user, or the business to the created report or web page Various methods are conceivable, such as being transmitted as link information by e-mail addressed to the user.

  In the client device 3, the control unit 30 receives the extraction result from the communication unit 36, outputs the received extraction result (step S304), and ends the process. The output of the extraction result in step S304 includes display on the display unit 33, transmission of print data from the communication unit 36 to the print medium, and print output thereby. Moreover, the output of the reading voice using the voice output unit may be included.

  Note that the processes in steps S105 to S110 and step S112 may be performed by batch processing each time a predetermined period of time elapses on the network community 100. The batch processing may be performed once a day, for example, in a time zone where the communication load and processing load are low, such as a time zone where transmission frequency is low. In this case, information indicating characteristics in association with the user identification information is stored in the storage unit 11 and updated by batch processing. When the text content S is received, the control unit 10 selects user identification information one by one, information indicating the characteristics of the transmission information stored in association with the selected user identification information, and the characteristics of the text content S. The similarity with the information indicating is calculated for all users. Thereby, for example, it is possible to reduce the processing load on the server device 1 by avoiding duplication of the derivation process of the information indicating the feature for each user that is performed each time the different text content S is received. Speeds up the response.

  In addition, the order in which the text content S is accepted (step S101) and the transmission information DB 112 is created is not limited to the example described above. In the flowchart of FIG. 5, the text content S is accepted after the transmission information DB 112 is created, that is, after the transmission information V is accumulated. However, the text content S may be received in advance and stored in the storage unit 11, and the control unit 10 may calculate the similarity with the text content S for the transmission information V transmitted thereafter.

(Example)
The embodiment described above will be described with specific examples.
FIG. 6 is an explanatory diagram showing an example of the content of the transmission information V. FIG. 6 shows transmission information (conversations on the bulletin board) V of users A, B, and C in the specific example of the network community 100 shown in FIG. 1 in time series. In the example shown in FIG. 6, “Pakchie” is listed as something recently eaten, and the transmission information V regarding “Pakchie”, which has recently become a boom, is shown.

  FIG. 7 is an explanatory diagram illustrating an example of the content of the text content S. The text content S is, for example, an article about “Pakchie”, and in particular, an article about the appeal of making and eating dishes using Pakchi at home. User D who is a business user who develops a unique product (for example, a cooking kit) using the ingredients of booming Pakuchi wants to ask a person who is interested in text content S to cooperate in product development. Suppose that By using the information processing system according to the present embodiment, user D can perform an operation of uploading text content S by selecting “extraction request” from the top page after logging in to network community 100 as a business user. That's fine.

When the text information S shown in FIG. 7 is uploaded in a state where the transmission information V group shown in FIG. 6 is recorded in the transmission information DB 112, the server apparatus 1 starts from the text content S (and the link destination) as follows. A list of frequent nouns and the number of appearances thereof is derived as information indicating characteristics.
(Pakchi, 25 times)
(Park, 10 times)
(Cultivation, 5 times)
(Cooking, 3 times)

Then, in the transmission information V group shown in FIG. 6, the transmission information V group that includes the user whose user identification information is “00003” (user name “userB”) as the sender (including other than the transmission information V shown in FIG. 6). ), The following list of frequent nouns and their number of appearances is derived as characteristic information.
(Pakchie, 110 times)
(Park, 25 times)
(Cultivation, 10 times)
(Cooking, 5 times)

  In the above example, frequent nouns are duplicated, and thus the similarity calculated by the cosine similarity based on the frequent nouns having the appearance count of 5 or more is “1”, for example.

  In the embodiment, the control unit 10 of the server device 1 creates display information (screen data) to be displayed on the client device 3 when transmitting the result of the extraction process in step S114 shown in the flowchart of FIG. As a result of the extraction process, the display information is transmitted (output) to the client apparatus 3. FIG. 8 is a flowchart illustrating an example of a creation process procedure in which the control unit 10 creates display information for displaying the extraction result.

  The control unit 10 narrows down, for example, the top 30 user identification information items sorted in descending order of similarity in step S113 in the procedure shown in the flowchart of FIG. 5 (step S401).

  The control unit 10 selects from the narrowed-down user identification information one by one in descending order of similarity (step S402), reads the user name of the selected user identification information from the user information DB 111, and displays the character information in the display information. (Or an image) is output (step S403). Next, the control unit 10 outputs numerical information indicating the similarity stored in association with the selected user identification information to the display information as character information (or an image) (step S404). Further, the control unit 10 selects a predetermined number of transmission information V in the transmission information V group transmitted by the selected user identification information as a main comment, and outputs it as display information (step S405).

  In step S405, the control unit 10 first selects transmission information V including more nouns identified as frequent nouns, and further selects transmission information V including more different frequent nouns. When the selected result is within the predetermined number, the control unit 10 may use the selected result as a main comment. When the selected result is more than the predetermined number, for example, the outgoing time is the latest outgoing information. You may make it select V preferentially. As described above, the control unit 10 gives priority to the extracted transmission information V according to the number of frequent nouns, the number of frequent nouns, and the transmission time, and selects only a predetermined upper number as main comments. Good.

  The control unit determines whether or not a predetermined number of user identification information has been selected from the top (step S406). If it is determined that the user identification information has not been selected (S406: NO), the process returns to step S402, and then similar User identification information with a high degree is selected.

  If it is determined in step S406 that the selection has been made (S406: YES), the control unit 10 outputs the display information as Web page data or image document data such as pdf (step S407), and ends the creation process. To do.

  FIG. 9 is an explanatory diagram illustrating a screen example displayed on the display unit 33 of the client device 3 based on the display information created by the control unit 10. The example shown in FIG. 9 corresponds to the result of extraction processing (S114) performed on the transmission information V on the network community 100 including the transmission information V group shown in FIG. 6 based on the text content S of FIG. And it is an example of the screen displayed based on the display information produced as a result of an extraction process by the procedure shown in the flowchart of FIG. As shown in the example of FIG. 9, on the screen displayed based on the display information, the user identification information is displayed in descending order of similarity, and the calculation result of similarity and the similarity are contributed for each user identification information. A plurality of main transmission information V including frequent nouns are included.

  As shown in the conversation example shown in FIG. 6, the user B whose user identification information extracted with the highest similarity in the example shown in FIG. 9 is “00003” (user name “userB”) It is a user who has an ambition to cook using food. Furthermore, from the conversation example shown in FIG. 6, the user B can later understand that “Today is a friend and my house!” And “My house”. That is, it can be said that the user is likely to be interested in the text content S of FIG. 7 and matches the person image imaged by the user D who is a business user. Thus, extraction of the user identification information of an appropriate person is realized by an easy operation of uploading the text content S that is likely to match the person image to be imaged. At this time, the user D does not need to perform an operation such as registration of a keyword corresponding to the attribute information.

  In the description of the flowchart of FIG. 5 and the description with reference to the specific examples of FIGS. However, the feature information derived based on the result of the morphological analysis is not limited to this. For example, as will be described later, information indicating a feature may be obtained by obtaining a vector including each component of a word including a related word by referring to a related dictionary in which the related word of the frequent noun is registered. Or it is good also as information which calculates TF-IDF (Term Frequency-Inverse Document Frequency) of a frequent noun, and shows the characteristic. Other information obtained by a method performed by a natural language processing technique may be used.

(Modification 1)
FIG. 4 shows an example in which the transmission information DB 112 stores the transmission information V in association with the transmission time. In the first modification, the transmission time is used in the extraction process performed by the control unit 10 of the server device 1. FIG. 10 is a flowchart illustrating another example of the procedure of the extraction process performed by the server device 1. Also, among the processing procedures shown in the flowchart of FIG. 10, procedures common to the processing procedures shown in the flowchart of FIG. Furthermore, in FIG. 10, since the processing procedure in the client apparatus 3 is the same, illustration and description are omitted.

  In the first modification, the control unit 10 of the server device 1 further transmits a call from the call information V group extracted from the call information DB 112 in step S107 in the most recent predetermined period based on the call time stored in association with it. The transmitted transmission information V group is narrowed down (step S121). The most recent predetermined period is, for example, three months from the time of uploading the text content S (the date and time accepted in step S101). Note that “most recent” is not necessarily strictly based on the time when the text content S is received, and should be interpreted in the sense of excluding past utterances.

  The control unit 10 assigns a weighting coefficient that becomes a higher numerical value in the time series based on the associated transmission time for each transmission information V included in the narrowed transmission information V group (step S122). . In step S122, for example, the control unit 10 assigns a coefficient of 1.0 to the transmission information V transmitted within the last one week and a coefficient of 0.9 to the transmission information V transmitted within the last one week to within two weeks. To do. A factor of 0.8 is given to outgoing information V sent within the last two weeks to three weeks, and a constant of 0.5 is assigned to outgoing information V sent within the last three weeks to three months. May be given.

In the derivation of the information indicating the characteristics in step S109, the control unit 10 multiplies the number of appearances of words, for example, nouns, obtained from the transmission information V by morphological analysis by the weighting coefficient given in step S122. A weight is applied to the total of For example, the transmission information V including “Pakchii” is sent once every one week within the last one week, within the last one week to two weeks, the last two weeks to three weeks, and the last three weeks to three months. If there is, the number of appearances is calculated as follows instead of “4”. Although the above-described example (1.0, 0.9, 0.8, 0, 5) is used as the coefficient, it is needless to say that the coefficient is not limited thereto.
(1 × 1.0 + 1 × 0.9 + 1 × 0.8 + 1 × 0.5) = 3.2 times As a result, the number of appearances is simply counted with the same weighting over the entire period after the transmission of the transmission information V is started. Rather than doing so, it is possible to calculate the degree of similarity by extracting words included in the latest outgoing information V as much as possible. In this case, it becomes possible to extract user identification information of a person who is likely to be interested in the text content S at the same time when the text content S is uploaded.

(Modification 2)
In the second modification, the control unit 10 of the server device 1 receives a plurality of text contents S and performs an extraction process. FIG. 11 is a flowchart illustrating another example of the procedure of the extraction process performed by the server device 1. Also, among the processing procedures shown in the flowchart of FIG. 11, the steps common to the processing procedure shown in the flowchart of FIG. Furthermore, in FIG. 11, since the processing procedure in the client apparatus 3 is the same, illustration and description are omitted.

  In the second modification, the control unit 10 of the server device 1 accepts a plurality of text contents S (step S131).

  In Modification 2 that accepts a plurality of text contents S, TF-IDF that derives feature words in a document used in natural language processing can be used in deriving information indicating the characteristics of text contents S in step S103. In this case, the control unit 10 may calculate the feature word and its TF-IDF using TF-IDF in the same manner when deriving information indicating the characteristics of the transmission information V extracted in step S109. The calculation of TF-IDF is not essential, and frequent nouns and the number of appearances may be derived for each of the plurality of text contents S, or the plurality of text contents S may be regarded as one text content S, The number of appearances may be derived.

  In the second modification, after storing the information indicating the characteristics of the transmission information V group for each user who is the caller derived in step S110, the control unit 10 includes the plurality of text contents S received in step S131, and the step The degree of similarity with the transmission information V group for each user stored in S110 is calculated (step S132).

  In step S132, the control unit 10 uses the TF-IDF derived from the plurality of text contents S and the TF-IDF derived from the transmission information V group for each user when using TF-IDF as described above. The cosine similarity may be calculated between them. Further, in step S132, when the control unit 10 derives information indicating the characteristics of each of the plurality of text contents S in step S103 without using TF-IDF, in step S132, transmission information for each user is obtained for each. The degree of similarity with group V is calculated. Further, when the control unit 10 derives information indicating the characteristics by regarding the plurality of text contents S as one text content without using the TF-IDF in step S103, the control unit 10 determines each user in step S132. One degree of similarity with the feature information derived from the transmission information group V is calculated.

  The plurality of text contents S may be text contents S having substantially the same content (theme), or may be text contents S having non-identical contents. For example, the plurality of text contents S having the same contents are a combination of an article about “Pak Chi” shown in FIG. 7 and another different article about eating “Pak Chi” at home. When using a plurality of text contents S having the same content, use TF-IDF to measure the similarity with the transmission information group V using only more important feature words, and calculate the similarity with high accuracy. Is possible.

  The text contents S whose contents are not the same are, for example, a combination of the article about “Pak Chi” and the article about “Thailand travel” shown in FIG. When using a plurality of text contents S whose contents are not identical, the control unit 10 regards the plurality of text contents S as one text content and derives frequent nouns and the number of appearances thereof to calculate the similarity, or The similarity may be calculated for each of the plurality of text contents S. When calculating the similarity for each of the plurality of text contents S, the control unit 10 may extract user identification information that is determined to be higher than a predetermined degree. This makes it possible to accurately narrow down and extract users who are interested in any of different articles.

  Thus, extraction processing that accepts a plurality of text contents S makes it possible to calculate the degree of similarity with higher accuracy and accurately extract user identification information of a user that matches a desired person image. It is expected that

(Modification 3)
In Modification 3, for each of the text content S and the transmission information V, the extraction process is performed by calculating the similarity using not only the words actually included (frequent nouns) but also related words. In Modification 3, the related dictionary is stored in the storage unit 11 or an external device, and can be read from the control unit 10. FIG. 12 is a flowchart illustrating another example of the procedure of extraction processing by the server device 1. Of the processing procedures shown in the flowchart of FIG. 12, the same steps as those shown in the flowchart of FIG. 5 are assigned the same step numbers, and detailed description thereof is omitted. Furthermore, in FIG. 12, since the processing procedure in the client apparatus 3 is the same, illustration and description are omitted.

  In the third modification, the control unit 10 of the server device 1 performs morphological analysis on the transmission information V group extracted from the transmission information DB 112 in step S107 (S108), and relates the words included in the transmission information V group. A word is extracted from the related dictionary (step S141). The control unit 10 derives information indicating the characteristics of the transmission information group V using not only words included in the transmission information group V obtained by morphological analysis but also related words (step S142). Regarding step S142, for example, the control unit 10 may extract related words as frequent nouns in association with the same number of appearances, or may include related words in feature words when calculating TF-IDF.

  Note that “related words” here include synonyms, co-occurrence words, and the like. For example, the related dictionary includes “Coriander”, “Kana”, etc., which are synonyms of “Pakchi” as related words. The related dictionary may include co-occurrence words “Thai cuisine”, “Thai”, and the like as related words of “Pak Chi”.

  Then, after storing the information indicating the characteristics of the transmission information V group (S110), the control unit 10 calculates the similarity between the text content S and the transmission information V group using the related words of the transmission information V group. The similarity calculated in association with the user identification information is stored in the storage unit 11 or the temporary storage unit 12 (S143). Specifically, in step S143, the control unit 10 vectorizes the frequent nouns that appear more than a predetermined number of times in the text content S as components, the frequent nouns that appear in the transmission information group V, and the related items. The cosine similarity is calculated with the word vectorized as each component. The Euclidean distance may be calculated.

  As for the related words, for the text content S received in step S103, the related words of the words included in the text content S are extracted from the related dictionary and used as information indicating the characteristics of the text content S. Good. The retrieval of related words may be performed not only on one of the text content S and the transmission information V group but on both.

  In this way, the extraction process using related words determines the degree of similarity in consideration of related words that are excluded from similar judgments even if they are essentially synonyms but are used differently. It is possible to calculate a higher degree of similarity when other terms are common. As a result, it is possible to accurately and accurately extract user identification information of a user that matches a desired person image.

  In addition, the above-described modified examples 1 to 3 can also be realized as a form in which any two or all of them are combined.

  In the above-described embodiment, the network community 100 is not limited to the participants such as the questionnaire community described above, but is widely disclosed such as Twitter (registered trademark), FaceBook (registered trademark), and blog. SNS may be used. Therefore, not only questionnaires on products / services, but also information that identifies the target person widely using a variety of transmission information including products / services, as well as impressions on the contents of television broadcasts and radio broadcasts, etc. Can also be extracted. At this time, information that is intentionally attached as a “tag” used in the SNS among the transmitted information may be extracted after a process of excluding it. This makes it possible to extract a person who is potentially a target regardless of the contents of the “tag”. Furthermore, the network community 100 is not only a place for a questionnaire survey related to product development, but may also be a place for information exchange in job hunting activities, job change activities, and other human resource matching services. By applying to job hunting activities, job hunting activities, and other human resource matching services, it is possible to extract users who match a desired person image envisioned by the personnel manager or the manager of the matching service.

  Moreover, in this indication, it was set as the structure which extracts an appropriate person with the transmission information via the communication medium on the network community 100 as mentioned above. However, the invention of the present application is not limited to this, and conversations by a plurality of persons are recorded by one sound collecting device, and the recorded information is converted into text, or the interview information is converted into text. It is also possible to apply. In this case, an identification information database for identifying the person who participated in the dialogue and the person who answered the interview is created, and the transmission information database is created using the text data of the remarks from each as the transmission information. Then, the process of the flowchart of FIG. 5 is performed using the text text content S corresponding to the desired person. As a result, it is possible to extract appropriate person identification information not only from the transmission information in the virtual space such as the network community 100 but also from the conversation in the real space or the transmission information in the paper medium.

  It should be understood that the embodiment disclosed above is illustrative in all respects and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the meanings described above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

1 Server device (extraction device)
10 control unit 11 storage unit 111 user information DB (sender identification information database)
112 Transmission Information DB (Transmission Information Database)
12 Temporary storage unit 13 Communication unit 2, 3 Client device 20, 30 Control unit 23, 33 Display unit 26, 36 Communication unit

Claims (8)

A reception unit for receiving text content corresponding to a desired person image;
A transmission information database in which a plurality of transmission information transmitted from a plurality of senders is recorded in association with identification information for identifying each transmission information, and a sender identification information for identifying each sender of the transmission information is recorded. A storage unit for storing the sender identification information database,
For each caller using the caller information database and caller identification information database, a calculation unit that calculates the similarity between the caller information group sent by the caller and the text content received by the receiving unit;
An extraction apparatus comprising: an extraction unit that extracts, from the sender identification information database, sender identification information of a sender whose similarity calculated by the calculation unit is equal to or greater than a predetermined degree. The extraction device according to claim 1, wherein the extraction unit sorts the caller identification information in descending order of corresponding similarity. For each caller identification information extracted by the extraction unit, a creation unit is further provided for creating display information for displaying information related to the calculation of the similarity and a plurality of transmission information contributing to the similarity. The extraction device according to claim 1, wherein: In the outgoing information database, the outgoing time of outgoing information is recorded in association with the outgoing information,
The calculation unit includes:
For each sender, extract the outgoing information group sent by the sender,
From the extracted transmission information group, narrow down the transmission information group transmitted in the period of the predetermined length from the latest,
The extraction device according to claim 1, wherein a similarity with the text content received by the reception unit is calculated using the narrowed down transmission information group. In the outgoing information database, the outgoing time of outgoing information is recorded in association with the outgoing information,
The calculation unit includes:
Each of the outgoing information extracted from the outgoing information database is given a weighting coefficient that becomes a higher numerical value in order from the time when the outgoing time is closer to the calculation time of the similarity,
Multiply the given weighting coefficient by the number of appearances of the word included in the transmission information,
The similarity with the text content received by the reception unit is calculated for each of the senders using the words included in the transmission information group transmitted by the senders and the number of appearances of the words. Item 2. The extraction device according to Item 1. The reception unit receives a plurality of text contents together,
2. The calculation unit according to claim 1, wherein the calculation unit calculates the similarity for each of the plurality of text contents, or calculates the similarity using information indicating characteristics derived from the plurality of text contents. The extraction device described. The calculation unit includes:
Calculating similarity using words included in each of the transmission information or the text content received by the receiving unit and related words extracted from related dictionaries in which related words related to the words are recorded. The extraction device according to claim 1, wherein A transmission information database in which a plurality of transmission information transmitted from a plurality of senders is recorded in association with identification information for identifying each transmission information, and a sender identification information for identifying each sender of the transmission information is recorded. A computer program that allows a computer that can read and write to the sender identification information database to extract the sender identification information,
In the computer,
Receiving text content corresponding to a desired person image;
Calculating the degree of similarity between the transmission information group transmitted by the sender and the received text content for each sender using the transmission information database and the sender identification information database;
A computer program for executing the step of extracting, from the sender identification information database, sender identification information of a sender whose calculated similarity is a predetermined degree or more.

Images