JP2015164008A - Analyzer and analyzing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an analyzer capable of performing segmentation of consumers and calculating the influences among consumers and segments to thereby provide information useful for efficient marketing at low cost.SOLUTION: The analyzer of speech information including the content of speech made by plural users and relationship information among the speeches, includes: a collection section that collects the speech information; an analyzing section that analyzes the frequency of appearance of a specific keyword in the speech content; a segmenting section that segments the plural users who made the speech on the content based on analysis of the frequency of appearance of the specific keyword; a graph generation section that creates a vector-graph among the plural users who made the speech based on the relationship information among the speech; and a calculation section that calculates the information on the relationship among the users based on the information on the segmented by the segmenting section; and the information relevant to the vector-graph created by the graph generation section.

Description

  The present invention relates to an analysis apparatus and an analysis method.

  In recent years, social media such as word-of-mouth sites, blogs, and social networking services (SNS) have been widely used as communication tools. The purpose of using social media as a communication tool is to interact with people with the same hobbies and preferences. Therefore, it is expected to collect and utilize information on personal interests and preferences and information on human relationships on social media from conversations conducted on social media.

  On the other hand, in marketing, for the purpose of determining the target purchase layer, consumer segmentation is performed to divide consumers into segments with similar needs and purchase behavior. By setting the target purchase layer where the company has an advantage and planning and executing marketing based on the needs and preferences of the target purchase layer, such as new product planning and campaign implementation, effective marketing measures can be performed at low cost. .

  On the other hand, for example, Patent Document 1 discloses a technique for segmenting consumers based on preferences based on analysis of questionnaire results. Patent Document 2 discloses a technique for creating a user interest correlation graph representing the similarity of interest between users from an input representing the relationship between a user and an item, such as a web content usage history, and segmenting the users. Yes.

JP 2008-293310 A JP 2011-141666 A

  It is possible to segment consumers (users) by using the techniques disclosed in Patent Documents 1 and 2. However, it is impossible to further narrow down marketing measures, such as which consumers in the segment are effective in conducting campaigns. In addition, it is not possible to predict to which segment other than the target segment.

  Therefore, an object of the present invention is to provide information useful for effective marketing at low cost by segmenting consumers and calculating the influence between consumers and the influence between segments.

  A representative analysis device according to the present invention includes a collection unit that collects the utterance information, and a specification in the utterance content in the utterance information analysis device that includes the utterance contents of a plurality of users and the relationship information between the utterances. An analysis unit that analyzes the appearance of the keyword, a segmentation unit that segments the plurality of users who have spoken the content based on the analysis of the appearance of the specific keyword, and the relationship information between the statements Based on the graph creation unit that creates a directed graph of the relationship between the plurality of users who have made a statement, information on the segmentation of the segmentation unit, and information on the directed graph creation of the graph creation unit, information on the relationship between the users is calculated And a calculating unit.

  The present invention is also grasped as an analysis method.

  According to the present invention, since consumers are segmented and the influence between consumers and the influence between segments are calculated, information useful for effective marketing at low cost can be provided.

It is a figure which shows the example of a structure of a consumer segmentation apparatus. It is a figure which shows the example of a structure of a consumer segmentation program. It is a figure which shows the example of accumulation | storage social media data. It is a figure which shows the example of user preference data. It is a figure which shows the example of the processing flow of a user preference analysis part. It is a figure which shows the example of user segment data. It is a figure which shows the example of the processing flow of a user segmentation implementation part. It is a figure which shows the example of social graph data. It is a figure which shows the example of the processing flow of a social graph preparation part. It is a figure which shows the example of the processing flow of a segment influence calculation part. It is a figure which shows the example of calculation of segment influence. It is a figure which shows the example of the processing flow of a segmentation result display part. It is a figure which shows the example of the user interface screen by the display of a user unit. It is a figure which shows the example of the user interface screen by the display of a segment unit. It is a figure which shows the example of a structure of the consumer segmentation program of Example 2. FIG. It is a figure which shows the example of the sales data of Example 2. FIG. It is a figure which shows the example of the processing flow of the sales data comparison part of Example 2. FIG.

  The preferred consumer segmentation is described below with reference to the drawings. FIG. 1 is a diagram illustrating an example of the configuration of a consumer segmentation device. The CPU 101 constituting the consumer segmentation device receives instructions from the consumer segmentation program stored in the main storage device 102, and reads, calculates, and writes data in the consumer segmentation program. The main memory device 102 stores data processed by the consumer segmentation device and the entire program. The main storage device 102 is, for example, a RAM. The auxiliary storage device 103 stores data processed by the consumer segmentation device and the entire program. The auxiliary storage device 103 is, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive).

  The input device 104 is a device that receives a command from the user to the consumer segmentation device and transmits it to the CPU 101. The input device 104 is, for example, a mouse or a keyboard. The output device 105 is a device that displays a user interface screen in the consumer segmentation device. The output device 105 is, for example, a liquid crystal display. The network interface 106 accesses the Internet and transmits and receives social media data. The network interface 106 is, for example, a LAN card. These devices are connected by a bus 107 and input / output data via the bus 107.

  FIG. 2 is a diagram showing an example of the configuration of a consumer segmentation program executed by the CPU 101 and data. In FIG. 2, a social media data collection unit 202, a user preference analysis unit 205, a user segmentation execution unit 207, a social graph creation unit 209, a segment influence calculation unit 211, and a segmentation result display unit 212 are programs executed by the CPU 101. is there. The accumulated social media data 203, word classification data 204, user preference data 206, user segment data 208, and social graph data 210 are data processed by a program of the consumer segmentation device. These data are stored in the main storage device 102 in FIG. 1, but may be stored in the auxiliary storage device 103 as necessary, such as the amount of data.

  Hereinafter, processing in each program will be described. The social media data collection unit 202 collects social media data 201 from the network via the network interface 106 and stores it as accumulated social media data 203. The accumulated social media data 203 is obtained by converting the social media data 201 collected by the social media data collection unit 202 into a format for storing in the apparatus. An example of the accumulated social media data 203 is shown in FIG.

  The speech ID 301 constituting the accumulated social media data 203 is an identifier composed of an arbitrary character string or numerical value for distinguishing each speech. The speaker ID 302 is an identifier for identifying the person who has made each statement. For example, when the social media data 201 is a microblog, the user name can be used. The link information 303 indicates the relationship between two statements. The link information 303 includes two types of link type and link source. The type of link is a character string representing how two statements are related. For example, when the social media data 201 is a microblog, it is “quotation of remark” or “reply to remark”. The link source indicates the message ID of another message related to the message. The comment date 304 indicates the date when the article was posted on social media. The comment content 305 is a text portion of an article posted on social media.

  Returning to the description of FIG. When the user preference analysis unit 205 starts processing according to conditions specified in advance, the user preference analysis unit 205 analyzes the accumulated social media data 203 in a specified period and generates user preference data 206. An example of the user preference data 206 is shown in FIG.

  User preference data 206 represents user preferences collected from social media. A user ID 401 constituting the user preference data 206 corresponds to the speaker ID 302 in the accumulated social media data 203. The preference keyword 402 is a category name when a word acquired from the stored social media data 203 or a word acquired from the stored social media data 203 is classified based on the word classification data 204. Here, the word classification data 204 is composed of a pair of a word and a category name to which the word belongs. For example, the word classification data 204 has a plurality of sets, each including a product name such as “vegetable juice” and a product category name such as “beverage” corresponding to the product. The preference keyword 402 corresponds to, for example, a specific product name, product category name, event name, and the like.

  The preference degree 403 is the appearance ratio of the preference keyword 402. The preference degree 403 represents the strength of the user's interest in the preference keyword 402. In the user preference data 206, one user ID 401 is associated with a set of one or more preference keywords 402 and a preference degree 403.

  Next, an example of the processing flow of the user preference analysis unit 205 is shown in FIG. When the user preference analysis unit 205 starts processing based on a condition specified in advance, the user preference analysis unit 205 acquires accumulated social media data 203 included in the specified period, and the utterance included in the acquired data is based on the speaker ID 302. Classification is performed for each user (step 501).

  After step 501, the user preference analysis unit 205 aggregates words included in each user's remarks in the accumulated social media data 203 and the appearance frequency of the words (step 502). For example, each utterance is divided into words using morphological analysis which is a well-known technique, and the total is performed.

  The user preference analysis unit 205 sets a preference keyword 402 after step 502 (step 503). For example, a word included in the accumulated social media data 203 is set as the preference keyword 402. At this time, only the word selected by the user may be set as the preference keyword 402. Further, by referring to the word classification data 204, the word and the category may be associated, and the category name to which the word belongs may be set as the preference keyword 402.

  The user preference analysis unit 205 calculates the preference degree 403 after step 503 (step 504). The preference degree 403 is calculated based on the appearance frequency of the preference keyword 402. For example, when the preference keyword 402 is a word, the preference level 403 is set as the appearance frequency of the word, and when the preference keyword 402 is a category, the preference level 403 is set as the appearance frequency of all words included in the category. Or it is good also considering what normalized the appearance frequency of the word with the appearance frequency of all the words contained in a user's utterance as the preference degree 403. FIG.

  Returning to the description of FIG. The user segmentation execution unit 207 analyzes the user preference data 206 to perform user segmentation, and generates user segment data 208. An example of the user segment data 208 is shown in FIG. The segmentation target user list 601 constituting the user segment data 208 is a set of user IDs 401 to be segmented. As a user who is the target of segmentation, for example, a user whose appearance frequency of a preference keyword is not zero may be included, or a user whose appearance frequency is zero may be included. Or it is good also as a user specified beforehand by a certain standard. The segment user list 602 is a set of user IDs 401 included in the same segment. The segment feature 603 represents the feature of the segment based on the user's preference included in the segment. The content of the segment feature 603 depends on the segmentation method. For example, when the segmentation method according to the preference level of the preference keyword “health” is used, the segment features are “health preference level is 0.4 or less” and “health preference level is 0.4 or more”. can do.

  Next, an example of the processing flow of the user segmentation execution unit 207 is shown in FIG. When the user segmentation execution unit 207 starts processing according to a predesignated condition, the user segmentation execution unit 207 acquires user preference data 206, and from among all users included in the user preference data 206, the segmentation target user is based on the preference keyword 402. Is extracted (step 701). For example, a user ID 401 including a specific preference keyword 402 such as “vegetable juice”, that is, a list of user IDs 401 whose preference level is not zero is acquired, thereby extracting a user who is a segmentation target. A user ID 401 that does not include the preference keyword 402 may be added, or a user that matches the condition based on a condition other than the preference keyword 402 that is set in advance and stored in the auxiliary storage device 103 or the main storage device 102. May be extracted. Participation in social media includes registration of gender, age range, region, etc., and may be used as a condition when such information is made public. The extracted segmentation target users are stored in the segmentation target user list 601 of the user segment data 208.

  After step 701, the user segmentation execution unit 207 performs user segmentation (step 702). The segmentation target user is divided into a plurality of segments by segmentation. The segmentation is performed by a known technique such as classification of users according to the preference level of a specific preference keyword and application of a cluster analysis method based on the preference keyword and preference level similarity of each user. For example, when segmentation is performed depending on whether the user is interested in a specific preference such as “health”, the user who performs the segmentation sets a threshold (for example, a value of 0.5) in advance for the preference level of the preference keyword “health”. By setting, the segment of the preference keyword “health” is divided into a segment composed of users having a threshold value or more and a segment composed of other users. Here, a segment having a preference level of zero may be specially provided.

  The user segmentation execution unit 207 calculates a segment feature 603 after step 702 (step 703). The segment feature 603 is calculated by a calculation method corresponding to the segmentation method. The calculation method is determined in advance by the user who performs the segmentation according to the segmentation method used for the user segmentation. As in the example described above, when segmentation is performed by setting a threshold value for the preference level of the preference keyword “health”, the segment feature 603 is “the health preference level is equal to or greater than the threshold value” by acquiring the set threshold value. , “Health preference is below threshold”. For example, when a plurality of preference keywords common to users included in the segment are acquired and the preference keyword is “summer clothes”, the segment feature 603 may be “summer clothes”.

  Returning to the description of FIG. The social graph creation unit 209 acquires the accumulated social media data 203 and the user segment data 208 and creates social graph data 210. An example of a social graph constituted by the social graph data 210 is shown in FIG.

  The social graph data 210 is a set of data constituting social graphs 821 and 822 including users 801 to 809 and directed edges 811 to 818. Social graphs 821 and 822 indicate the relationship between the users included in the segments 831 and 832 and other users. For example, the social graph 821 represents the relationship between the users 801, 802, and 803 included in the segment 831 and other users. When the social media data 201 is a microblog, the social graph 821 represents a conversation exchange from the users 801, 802, and 803 included in the segment 831.

  The users 801 to 809 correspond to the user ID 401 and configure social graph points (nodes). The directed sides 811 to 818 express the presence / absence of the relationship between users, and each directed side 811 to 818 has a weight which is a numerical value indicating the strength of the relationship between users. For example, when the social media data 201 is a microblog, the directed side 811 corresponds to whether or not a conversation is transmitted from the user 801 to the user 802, and the weight corresponds to the transmission frequency of the conversation from the user 801 to the user 802.

  Next, an example of the processing flow of the social graph creation unit 209 is shown in FIG. The social graph creation unit 209 selects one segment from the user segment data 208 when the process is started under the conditions specified in advance (step 901). Subsequently, after step 901, the social graph creation unit 209 extracts the user's remarks included in the selected segment from the accumulated social media data 203 (step 902).

  After step 902, the social graph creation unit 209 counts the number of links between users included in the segment from the extracted user's comments (step 903). For example, when the social media is a microblog, for each user included in the target segment, for the reply made by the user, the list of reply destination users and the number of replies to each reply destination user are tabulated. The list of users who quoted the utterances and the number of times the user's citations were quoted are tabulated.

  After step 903, the social graph creation unit 209 determines the presence and weight of the directed sides 811 to 818 of the social graphs 821 and 822 based on the total number of links between users (step 904). Below, the case where the social media is a microblog will be described as an example, and the determination processing of the presence / absence of a directed edge and the weight between the user 801 and the user 802 will be described. First, when the user 802 replies to the user's 801 utterance, since the conversation is being sent from the user 802 to the user 801, the directed side 812 from the user 802 to the user 801 is changed. The number of replies is set as the weight of the directed side 812. In addition, when the user 802 quotes the remarks of the user 801, since the conversation is transmitted from the user 801 to the user 802, the directed side 811 from the user 801 to the user 802 is set, and the number of citations is set. The weight of the directed side 811 is used.

  In this example, the social graph is configured based on the conversation directly transmitted from the user included in the segment, such as “quoting the remarks of the user 801 by the user 802”, but “the remarks of the user 801 quoted by the user 802 is A social graph including indirect influences may also be created by targeting replies from the callee user or quotes from the callee user, such as “when the user 804 quotes”.

  The above processing is executed for all segments, and social graphs 821 and 822 for each segment are created (step 905).

  And the social graph preparation part 209 produces the social graph of the whole segmentation object user after step 905 (step 906). That is, by executing steps 901 to 904 for the entire segmentation target user, a social graph of the entire segmentation target user is created.

  Returning to the description of FIG. The segment influence calculation unit 211 reads the user segment data 208 and the social graph data 210 and calculates the segment influence. The segment influence is the magnitude of the influence each segment has on other users. The segment influence is calculated based on the social graph of the segment.

  Next, an example of the processing flow of the segment influence calculator 211 is shown in FIG. When the segment influence calculation unit 211 starts the process under a condition specified in advance, the segment influence calculation unit 211 selects one segment for calculating the segment influence (step 1001). Subsequently, after step 1001, the segment influence calculation unit 211 acquires the social graphs 821 and 822 of the selected segment from the social graph data 210 (step 1002).

  After step 1002, the segment influence calculation unit classifies the directed edges 811 to 818 of the acquired social graphs 821 and 822 according to the user pointed to by the directed edges 811 to 818 (step 1003). For example, when classifying the directed edge 811 of the social graph 821, the user 802 to which the directed edge 811 points is “user included in the same segment as the user 801 that is the transmission source”, “transmission source” It is classified according to whether it belongs to “segmentation target user not included in the same segment as user 801” or “user not segmentation target user”. Further, it may be classified into “users who send a conversation to the user 801 who is a transmission source” and “users who do not send a conversation to the user 801 who is a transmission source”.

  In this example, the user 802 is a user included in the same segment as the user 801, and in order to send a conversation to the user 801, the directed side 811 is “the conversation of the user included in the same segment as the transmission source”. Can be classified as “outgoing”. Further, the user 803 can be classified as “a user who is included in the same segment as the user 801 who is the transmission source and does not send a conversation to the user 801”, and is registered in the segmentation target user list 601 shown in FIG. The user 804 who has not been classified can be classified as “a user who is not a segmentation target user”. This means that conversations are classified based on the preference of the user who is the destination of the conversation.

  After step 1003, the segment influence calculation unit 211 changes the weights of the directed sides 811 to 818 according to the classification of the directed sides 811 to 818 (step 1004). For example, when the classification of the directed edges 811 to 818 is “transmission of conversations to users included in the same segment” for the weights of the directed edges 811 to 818, a predetermined numerical value P1 is set. Multiply by “sending conversation to segmentation target users not included in the same segment”, multiplying by a predetermined numerical value P2 and “sending conversation to users who are not segmentation target users” Multiplies a predetermined numerical value according to the classification of the directed edges 811 to 818 to correct the weight of the directed edge, such as multiplying by a predetermined numerical value P3. This means that the weights of the directed sides 811 to 818 are corrected according to the purchase possibility of the user who is the destination of the conversation.

  The segment influence calculation unit 211 calculates the segment influence after step 1004 (step 1005). For the social graphs 821 and 822 whose weights are corrected in step 1004, the segment influence is calculated according to the example shown in FIG. For example, when the segment influence of the segment 831 shown in FIG. 8 is obtained, the sum of the weights corrected in step 1004 is calculated for all the directed edges 811 to 816 included in the social graph 821 and included in the segment 831. Divide by 3 which is the number of users. The segment influence is obtained by normalizing the sum of the weights of the directed sides 811 to 818 from the users included in the segment by the number of users included in the segment in the social graph. For this reason, when a user included in a segment actively sends a conversation to another user, the influence of the segment increases.

  By performing the above processing for all segments, the segment influence of each segment is calculated (step 1006). By obtaining the segment influence, it becomes possible to grasp the influence of the user included in the segment.

  Returning to the description of FIG. The segmentation result display unit 212 displays the segmentation result on the user interface screen based on the processing result in the segment influence calculation unit 211. FIG. 12 shows an example of the processing flow of the segmentation result display unit 212, and FIGS. 13A and 13B show examples of user interface screens by the segmentation result display unit.

  A user interface screen 1301 illustrated in FIG. 13A includes a segment display unit 1302, a social graph display unit 1303, and a social graph display change UI 1304. The segmentation result display unit 212 displays the user segment data 208 and the segment influence on the segment display unit 1302 of the user interface screen 1301 in a list format when the process is started under a condition specified in advance (step 1201). For example, the user segment data 208 and the segment influence are displayed in a table format as shown in FIG. 13A. The segment name display unit 1311 constituting the segment display unit 1302 displays a character string for associating with the social graph displayed on the social graph display unit 1303. Further, the segment influence display unit 1312 displays the segment influence calculated by the segment influence calculation unit 211, and the segment feature display unit 1313 displays the segment feature 603 in the user segment data 208.

  As a result, it is possible to identify a segment having influence on other users based on the display of the segment influence display unit 1312 and grasp characteristics relating to the preference of the segment based on the display of the segment feature display unit 1313. it can. In the example shown in FIG. 13A, it can be seen that “Segment A” has a higher segment influence than other segments and is composed of users who have a strong interest in “health”. Based on this, users included in “Segment A” by implementing marketing measures that attract the interest of users who make up “Segment A”, such as distributing advertisements that appeal to health targeting “Segment A” A high effect is obtained with respect to the whole, and an effect due to an influence from a user included in “Segment A” can be expected for users other than “Segment A”.

  After Step 1201, the segmentation result display unit 212 displays the social graph of the entire segmentation target user on the social graph display unit 1303 of the user interface screen 1301 (Step 1202). For example, the users 801 to 809 and the directed sides 811 to 818 in the social graph of the entire segmentation target user are displayed as the users 1331 to 1336 and the directed sides 1341 to 1347 shown in FIG. 13A. The directed sides 1341 to 1347 are displayed by changing the thickness of the line according to the weights of the directed sides 811 to 818 of the social graph. As a result, for example, a social graph as shown in the display 1321 is displayed.

  After step 1202, the segmentation result display unit 212 determines the social graph to be displayed from the social graphs of the segments included in the user segment data 208 based on the user segment data 208 and the segment influence (step 1203). . For example, a threshold is set for the segment influence in advance, and a social graph of a segment having a segment influence greater than the set threshold is set as a display target. Further, the threshold value is configured to be changeable by the user using the input device 104 by a segment display reference changing unit 1361 described later.

  After step 1203, the segmentation result display unit 212 displays the segment social graph on the social graph display unit 1303 (step 1204). For example, the social graph of the segment to be displayed in step 1203 is displayed by the following process. First, in the social graph of the entire segmentation target user displayed in step 1202, the users included in the segment are highlighted. For example, a user included in a segment to be displayed is surrounded by a frame 1351, and the segment name 1311 in the segment display unit 1302 is displayed together with the frame 1351 like a character string 1352. Further, highlighting is performed by changing the color of the side belonging to the social graph of the segment, such as directed sides 1341, 1342, and 1343. The above processing is performed for each segment to be displayed. At that time, the social graphs of different segments are displayed so as to be identifiable, such as displaying the social graphs of different segments in different colors. Thereby, it can be determined whether or not each displayed segment is influenced by other segments, and the segment that is the source of conversation can be determined as the target purchase layer.

  In the example of FIG. 13A, it can be seen from the display content in the social graph display unit 1303 that “segment A” has an influence on many users including users included in “segment B”. Further, it can be seen from the display contents in the segment display unit 1302 that “Segment A” is composed of users who are strongly interested in “health”. For all the users included in “Segment A” by conducting marketing measures that attract the interest of “Segment A” users, such as the distribution of health-oriented advertisements targeting “Segment A”, which has high segment influence A high effect is obtained, and an effect due to an influence from a user included in “Segment A” can be expected for other users including “Segment B”.

  It can also be seen that among the users included in “Segment A”, “User Aaa” is the source of conversations with users included in other segments. In addition to marketing measures targeting “Segment A”, marketing measures such as separate advertisement distribution are targeted at users who are the source of conversation with users included in other segments such as “User Aaa” As a result, the influence of the user included in “Segment A” on other users can be strengthened.

  After step 1204, the segmentation result display unit 212 receives an instruction from the user (user performing the segmentation) of the consumer segmentation device from the input device 104, and updates the display of the segmentation result (step 1205). For example, an instruction from a user of the consumer segmentation device is transmitted to the segmentation result display unit 212 through the social graph display change instruction UI 1304.

  The segment display reference change UI 1361 constituting the social graph display change instruction UI 1304 receives an instruction from the user of the consumer segmentation device, and changes the segment influence threshold in the social graph display. For example, as shown in the input field 1361 of FIG. 13A, it is assumed that an instruction from the consumer segmentation device user can input a threshold value and a condition. Thereby, the user of the consumer segmentation device can select how high the social graph of the segment having high influence is to be displayed.

  In addition, the display segment change UI 1362 receives an instruction from the user of the consumer segmentation device and switches between displaying and hiding the social graph of the segment. For example, as shown in a selection field 1362 in FIG. 13A, the consumer segmentation device user can select display / non-display. For example, in the case of marketing planning for an in-house product, consider the case where the purpose is to acquire a new user from a competitor's product user. First, a user group who is interested in the company's products is segmented as a segmentation target user, and a segment with high segment influence is considered. If these highly influential segments are “Segment A” and “Segment B”, “Segment A” and “Segment B” are displayed, and “Segment C” that is interested in competing products from other companies. "To be displayed together with" Segment A "and" Segment B ". Then, targeting any of the segments that have many users in “Segment C” of “Segment A” and “Segment B”. A user can be newly acquired as a user of the company's product.

  The graph display unit change UI 1363 receives an instruction from the user of the consumer segmentation device, and a social graph of the user unit with the user shown in the display 1321 as a point of the graph and the segment shown in the display 1391 of FIG. The display of the social graph in units of segments is switched, and either one is displayed on the social graph display unit 1303. For example, as shown in the graph display unit change UI 1363 in FIG. 13A, it is assumed that the user of the consumer segmentation device can select whether to display a social graph in units of users or a social graph in units of segments.

  In the display 1391 of the social graph in segment units, the segments specified by the display segment changing unit 1362 are displayed as graph points 1371 and 1372. In addition, reception / transmission of conversation between segments is displayed as a directed side 1381. The directed side 1381 is displayed by changing the thickness of the line based on the weight of the related side in the social graph. For example, in the social graph data 821 and 822, the sum of the weights of the sides facing the user included in the segment 1372 from the user included in the segment 1371 is calculated, and the thickness of the line is determined based on the sum of the weights. This makes it possible to visualize a social graph that expresses the relationship between segments, and it is easy to identify segments that have influence as a source of conversation. As in the example already described, by distributing the advertisement to the segment that is the transmission source of the conversation, the effect of the advertisement transmission can be obtained for many users.

  As described above, user preferences can be extracted from social media to segment users, and information about influential users and segments can be provided. This enables effective marketing such as distributing advertisements to influential users and segments.

  In the second embodiment, an example of a consumer segmentation method and apparatus capable of utilizing sales data information in addition to information from social media data will be described. The user of the consumer segmentation device according to the second embodiment can identify a segment related to an accidental change in sales data and use it for analysis of a change factor.

  FIG. 14 is a diagram illustrating a configuration example of a program and data executed by the CPU 101 in FIG. 1 in the second embodiment. In FIG. 14, the programs and data having the same reference numerals as those in FIG. In the consumer segmentation program of the second embodiment, sales data 1401 and a sales data comparison unit 1402 are added to the configuration of the consumer segmentation program shown in FIG. 2 of the first embodiment.

  Sales data 1401 is data describing sales for each product period, and an example of sales data is shown in FIG. The product name 1501 constituting the sales data 1401 is the name of the product handled by the company. The sales amount 1502 indicates the sales amount of the product described in the product name 1501 within a specific period, and includes, for example, daily or monthly sales.

  Returning to the description of FIG. The sales data comparison unit 1402 identifies a segment related to a change in sales based on the segment influence obtained by the segment influence calculation unit 211 and the sales data 1401. The sales data comparison unit 1402 is configured to be executed after step 1204, for example. An example of the processing flow of the sales data comparison unit 1402 is shown in FIG.

  First, the sales data comparison unit 1402 collects sales data 1401 for a certain period and detects a product in which a large change in sales has occurred (step 1601). For example, for each day, the current day's sales and the previous day's sales are compared, and a product whose sales have increased or decreased beyond the fluctuation range set by the user performing the segmentation is detected.

  After step 1601, the sales data comparison unit 1402 transmits the date on which the change in sales is detected and information on the product in which the change has occurred to the user preference analysis unit 205 and the user segmentation execution unit 207, thereby Segmentation is performed (step 1602). The sales data comparison unit 1402 transmits the date when the change in sales is detected to the user preference analysis unit 205. The user preference analysis unit 205 obtains accumulated social media data 203 from a date when a change in sales is detected to a certain period before, and creates user preference data 206. In addition, the sales data comparison unit 1402 transmits to the user segmentation execution unit 207 the product name for which the change in sales has been detected. In step 702, the user segmentation execution unit 207 extracts, as a segmentation target user, a user who has, as a preference keyword, a product name or product category in which a change in sales has been detected. User segment data 208 is created by performing segmentation for these users.

  After step 1602, the sales data comparison unit 1402 causes the segment influence calculation unit 211 to calculate the segment influence of each segment (step 1603). After step 1603, the sales data comparison unit 1402 causes the segmentation result display unit 212 to display the segmentation result (step 1604). Thus, an influential segment can be identified from among users who are interested in a product in which a change in sales has occurred immediately before the date on which the change in sales has been detected. Then, by paying attention to the content of the user's remarks included in the identified segment, it becomes possible to analyze the factors that change the sales. For example, immediately before a change in sales is detected, an influential segment is identified, and keywords that frequently appear in the remarks of users included in the identified segment are extracted. You can get it.

  As described above, by associating the actual sales change of the product with the user's social media remarks, it is possible to provide information on the segments that have an influence on the sales change. It is also possible to extract keywords that have an impact on sales fluctuations. This enables effective marketing that includes not only the influence on the remarks in social media but also the influence on the actual sales.

  In the above description, the case where the data collected from the outside is the social media data 201 collected through the Internet, the graph is the social graph data 210, and the case where the time series data is the sales data 1401 has been described. These examples are not intended to limit the scope of the claims.

201: Social media data, 202: Social media data collection unit, 203: Accumulated social media data, 204: Word classification data, 205: User preference analysis unit, 206: User preference data, 207: User segmentation execution unit, 208: User Segment data, 209: Social graph creation unit, 210: Social graph data, 211: Segment influence calculation unit, 212: Segmentation result display unit, 1401: Sales data, 1402: Sales data comparison unit

Claims (14)

In a speech information analyzing apparatus including the content of a plurality of user speeches and the relationship information between the speeches,
A collection unit for collecting the remark information;
An analysis unit for analyzing the appearance of a specific keyword in the content of the statement;
A segmenting unit for segmenting the plurality of users who have spoken the content based on an analysis of the appearance of the specific keyword;
A graph creation unit for creating a directed graph of the relationship between the plurality of users who have made a statement based on the relationship information between the statements;
A calculation unit for calculating information on a relationship between users based on information on segmentation of the segmentation unit and information on directed graphing of the graph creation unit;
The analysis apparatus characterized by having. The specific keyword is a preference keyword or a word related to the preference keyword,
The segmenting unit includes:
Calculating a preference degree representing the strength of each user's preference based on the appearance frequency of the preference keyword or the word related to the preference keyword included in the content of the remark;
The analysis apparatus according to claim 1, wherein the plurality of users are segmented based on the calculated preference degree. The utterance information is social media, and the relationship information between the utterances includes citations and the number of citations,
The graphing unit converts the citation between the statements into a directed graph,
The analysis device according to claim 1, wherein the calculation unit calculates influence between users based on the number of the directed graphs.   The said calculating part calculates the influence between users based on whether it is the directed graph between the users in the same segment, or the directed graph between the users in a different segment. Analysis device. Displays instructions for selecting a specific segment from the segments,
Select a segment according to the displayed instruction content,
A display unit for displaying a graphic representing the range of the selected segment, a graphic representing a user included in the selected segment, and a graphic representing a directed graph representing a user relationship included in the selected segment; The analysis apparatus according to claim 1, comprising: Displays instructions for selecting a specific segment from the segments,
Select a segment according to the displayed instruction content,
A directed graph representing user relationships included in the selected segment is collected between the selected segments;
The analysis apparatus according to claim 1, further comprising a display unit configured to display a graphic representing the selected segment and a graphic representing a directed graph between the collected segments. A comparison unit that collects time-series data, detects a change in the collected time-series data, notifies a date and time of the detected change to the analysis unit, and notifies a keyword related to the detected change to a segmentation unit Further comprising
The analysis unit analyzes a predetermined period before the notified date and time,
The analysis device according to claim 1, wherein the segmenting unit performs segmentation based on the notified keyword. A method for analyzing utterance information including the content of utterances of a plurality of users and relation information between the utterances,
A collecting step of collecting the remark information;
An analysis step of analyzing the appearance of a specific keyword in the content of the statement;
A segmentation step of segmenting the plurality of users who have spoken the content based on an analysis of the occurrence of the specific keyword;
A graph creation step for creating a directed graph of the relationship between the plurality of users who have made a statement based on the relationship information between the statements;
A calculation step for calculating information on a relationship between users based on information on segmentation of the segmentation unit and information on directed graphing of the graph creation unit;
The analysis method characterized by having. The specific keyword is a preference keyword or a word related to the preference keyword,
The segmentation step includes
Calculating a preference degree representing the strength of each user's preference based on the appearance frequency of the preference keyword or the word related to the preference keyword included in the content of the remark;
The analysis method according to claim 8, wherein the plurality of users are segmented based on the calculated preference degree. The utterance information is social media, and the relationship information between the utterances includes citations and the number of citations,
The graphing step converts the citation between the statements into a directed graph,
The analysis method according to claim 8, wherein the calculating step calculates influence between users based on the number of the directed graphs.   9. The influence between users is calculated based on whether the calculating step is a directed graph between users in the same segment or a directed graph between users in different segments. Analysis method. Displays instructions for selecting a specific segment from the segments,
Select a segment according to the displayed instruction content,
A display step for displaying a graphic representing the range of the selected segment, a graphic representing a user included in the selected segment, and a graphic representing a directed graph representing a user relationship included in the selected segment; The analysis method according to claim 8, further comprising: Displays instructions for selecting a specific segment from the segments,
Select a segment according to the displayed instruction content,
A directed graph representing user relationships included in the selected segment is collected between the selected segments;
The analysis method according to claim 8, further comprising a display step of displaying a graphic representing the selected segment and a graphic representing a directed graph between the collected segments. Further comprising a comparison step of collecting time series data and detecting a change in the collected time series data;
The analysis step analyzes a predetermined period before the date and time of the detected change,
9. The analysis method according to claim 8, wherein the segmenting step performs segmentation based on a keyword related to the detected change.

Images