JP2016045900A - Program, device and method detecting influencer having high influence degree in comment sentence to be submitted - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a program, a device and a method which can detect influencer having a high influence degree in a comment sentence to be submitted.SOLUTION: A program comprises: mental state determination means which outputs a mental state based on psychological keyword included in a comment sentence; score storage means which stores a score of the mental state for each quotation address; score update means which increases the score of the mental state of the reference comment sentence including the quotation address when the mental state is changed between the reference comment sentence and one or more forward comment sentences in the past, and/or increases the score of the mental state of the backward comment sentence when the mental state is changed between the reference comment sentence and one or more backward comment sentences in the future; and influencer detection means which outputs a prescribed number of quotation address in order of upper scores as influencer having a high influence degree for each mental state.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a technique for analyzing comment text posted to a communication site server such as SNS (Social Networking Service).

  In recent years, an unspecified number of third parties can actively transmit their own comment sentences (text information) via the SNS site server. The SNS site server publishes comment text posted by one user to a large number of users in a group. For example, there are facebook (registered trademark), twitter (registered trademark), google + (registered trademark), and mixi (registered trademark), which are also generally referred to as miniblog sites.

By analyzing the comment texts of a large number of users, it is possible to know the tendency according to the user's profile (age, gender, hobby, etc.). This is because the content of the comment text differs depending on the user profile even for the same product service.
On the other hand, there is also a technology that collects comment texts of a specific user in the SNS retrospectively and estimates the user's profile from the characteristics of words appearing in the comment texts (see, for example, Patent Document 1).

  In addition, there is a technique for efficiently extracting influencers based on other users' reactions (for example, the number of citations and the number of replies) to the posted content of each user using SNS (see, for example, Patent Document 2). The “influencer” means a user who has a large response from other users depending on the posted content. The user corresponding to the influencer means that the posted content introducing the product service is more effective as marketing or promotion.

  As another conventional technique, there is a technique for analyzing an advertisement effect using an AIDMA model that is an economics model relating to purchasing behavior (see, for example, Patent Document 3). According to the general knowledge of economics, the psychological state regarding the purchase behavior of the user is A (Attention)-> I (Interest)-> D (Desire)-> M (Memory) -> Change like A (Action). Each AIDMA phase is assigned to each type of advertisement or action (TV advertisement, magazine advertisement, email, Web page, banner click, bookmark, coupon acquisition, etc.), and the number of those (number of TV advertisements, banners) If it is a click, the number of clicks) is measured. This quantifies the impact of advertising on each phase of AIDMA.

JP 2013-196070 A JP 2014-021889 A JP 2003-044738 A

  In many cases, the comment text transmitted by the user includes the user's own psychological state consciously or unconsciously. In particular, according to the marketing application, it is desired to analyze the user's psychological remarks regarding the product service.

However, according to the technique described in Patent Document 1, the estimated user profile information is fixed and is not information that changes psychologically in time series.
Moreover, according to the technique described in Patent Document 2, what influence did the contribution content of the influencer have on the purchasing psychology of the merchandise service for other users? I can't judge until.
In addition, according to the technique described in Patent Document 3, the comment sentence posted to the SNS is not assumed for the estimation of the psychological state in the AIDMA model.

  In contrast, the inventors of the present application estimate the transition of the user's purchasing psychological state from the comment text posted to the SNS for a specific product service, and the user address (for example, the SNS account name) that contributed to the transition. Can you estimate the reply address)? I thought. For example, for each user address, can a score value contributing to the transition of the psychological state of another user for the product service be accumulated? I thought. This makes it possible to detect which user has an influence on the psychological state transition in marketing and promotion with respect to other users.

  Therefore, an object of the present invention is to provide a program, an apparatus and a method program, an apparatus and a method capable of detecting an influencer having a high influence level of a comment text to be posted.

According to the present invention, from a group of comment sentences transmitted for each contributor, a program for causing a computer mounted on the apparatus to function so as to analyze the effect of the user's psychological transition on the target keyword,
Psychological keyword dictionary means for registering psychological keywords in advance for each psychological state;
For each contributor, comment sentence acquisition means for acquiring comment sentences including the target keyword in time series,
For each contributor, when the psychological keyword of the psychological keyword dictionary means is included in the comment text, the psychological state determining means for outputting the psychological state of the psychological keyword,
Score storage means for storing a score of each psychological state for each citation address;
In the case of a reference comment sentence including a quote address in a group of comment sentences, if the psychological state changes between the reference comment sentence and one or more forward comment sentences in the past, the psychological state of the reference comment sentence And / or if the psychological state changes between the reference comment sentence and one or more backward comment sentences in the future, the psychological state score of the backward comment sentence is incremented. Score update means for instructing the score storage means;
The score storage means is used to cause the computer to function as influencer detection means for outputting a predetermined number of cited addresses from the top of the score as influencers having a high influence level for each psychological state.

According to another embodiment of the program of the present invention,
Target keywords are keywords based on product services,
It is also preferable that the computer further functions so that the quote address is a user address in SNS (Social Networking Service).

According to another embodiment of the program of the present invention,
Psychological keyword dictionary means that psychological state corresponds to AIDMA (Attention: attention-s> Interest: interest-> Desire: desire-> Memory: memory-> Action: behavior) model. It is also preferable to make the computer function so as to register psychological keywords for the keywords.

According to another embodiment of the program of the present invention,
The psychological state is assigned a numerical value that increases in the order of A->I->D->M-> A,
A time series of psychological transitions is a time series of psychological values.
The score updating means preferably causes the computer to function so that the numerical values arranged in time series are set to a moving average value within a predetermined range.

According to another embodiment of the program of the present invention,
It is also preferable that the score updating means sets the weight wb for the psychological state of the reference comment sentence and the weight wa for the psychological state of the backward comment sentence as different values, and causes the computer to function so that the score is weighted and incremented. .

According to another embodiment of the program of the present invention,
The score updating means preferably causes the computer to function such that the weight wb for the psychological state of the reference comment sentence and the weight wa for the psychological state of the backward comment sentence are incremented with different weights for each psychological state.

According to another embodiment of the program of the present invention,
The weight wb of the psychological state AIDM (Attention: attention-s> Interest: interest-> Desire: desire-> Memory: memory) is larger than the weight wb of the psychological state A ′ (Action: action)
The weight wa of the psychological state AIDM (Attention: attention-s> Interest: interest-> Desire: desire-> Memory: memory) is set in advance so as to be smaller than the weight wa of the psychological state A ′ (Action: action). It is also preferable to make the computer function.

According to another embodiment of the program of the present invention,
The score update means
When the psychological state changes between the reference comment sentence and one or more backward comment sentences in the future, the weight wa becomes smaller as the posting time interval between the reference comment sentence and the backward comment sentence becomes longer. It is also preferred to have the computer function to set.

According to the present invention, from a group of comment sentences transmitted for each contributor, an apparatus for analyzing the effect of a user's psychological transition on a target keyword,
Psychological keyword dictionary means for registering psychological keywords in advance for each psychological state;
For each contributor, comment sentence acquisition means for acquiring comment sentences including the target keyword in time series,
For each contributor, when the psychological keyword of the psychological keyword dictionary means is included in the comment text, the psychological state determining means for outputting the psychological state of the psychological keyword,
Score storage means for storing a score of each psychological state for each citation address;
In the case of a reference comment sentence including a quote address in a group of comment sentences, if the psychological state changes between the reference comment sentence and one or more forward comment sentences in the past, the psychological state of the reference comment sentence And / or if the psychological state changes between the reference comment sentence and one or more backward comment sentences in the future, the psychological state score of the backward comment sentence is incremented. Score update means for instructing the score storage means;
It is characterized by having influencer detection means for outputting a predetermined number of cited addresses from the top of the score as influencers having a high influence level for each psychological state using score storage means.

According to the present invention, using a device, a method of analyzing the effect of a user's psychological transition related to a target keyword from a group of comment sentences transmitted for each contributor,
The device
For each psychological state, a psychological keyword dictionary part in which psychological keywords are registered in advance,
For each citation address, it has a score storage unit that stores the score of each psychological state,
The device
For each contributor, a first step of acquiring comment sentences including the target keyword in time series;
For each contributor, if the comment text includes a psychological keyword in the psychological keyword dictionary, a second step of outputting the psychological state of the psychological keyword;
In the case of a reference comment sentence including a quote address in a group of comment sentences, if the psychological state changes between the reference comment sentence and one or more forward comment sentences in the past, the psychological state of the reference comment sentence And / or if the psychological state changes between the reference comment sentence and one or more backward comment sentences in the future, the psychological state score of the backward comment sentence is incremented. A third step for instructing the score storage means;
And a fourth step of outputting, for each psychological state, a predetermined number of cited addresses from the top of the score as influencers having a high influence level, for each psychological state.

  According to the program, apparatus, and method of the present invention, it is possible to detect an influencer having a high influence level of a comment text to be posted.

It is a system configuration diagram in the present invention. It is a functional block diagram of the psychological transition analyzer in this invention. It is explanatory drawing showing the comment text transmitted in time series. It is explanatory drawing which divided the comment text containing a target keyword for every user address. It is explanatory drawing which provided the psychological state to the comment text containing a target keyword. It is explanatory drawing showing the 1st score update in this invention. It is explanatory drawing showing the 2nd score update in this invention. It is explanatory drawing showing the score update which attached the 1st weight in this invention. It is explanatory drawing showing the score update which attached the 2nd weight in this invention.

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

  FIG. 1 is a system configuration diagram according to the present invention.

  According to FIG. 1, the psychological transition analysis apparatus 1 of the present invention is connected to the Internet. The psychological transition analysis device 1 communicates with the SNS site server 2 via the Internet. The psychological transition analysis apparatus 1 can obtain comment text for each contributor via the API (Application Programming Interface) from the SNS site server 2. The API is a rule interface for using the function of the application service, and is prepared as different for each of various servers.

  In addition, the psychological transition analysis apparatus 1 may accumulate comment texts in a database in advance. That is, the psychological transition analysis device 1 does not necessarily communicate with the SNS site server 2.

  An unspecified number of third parties can send comment text to the SNS site server 2 via the Internet using their own terminals 3. Moreover, the psychological transition analysis apparatus 1 analyzes the user psychological transition regarding a target keyword from the group of comment sentences transmitted for each contributor. The “target keyword” is a keyword based on the product service. In addition, a large number of comment sentences include those that include or do not include user addresses of other users (for example, account names, reply addresses, and comment addresses in SNS).

  FIG. 2 is a functional configuration diagram of the psychological transition analysis apparatus according to the present invention.

  The psychological transition analysis apparatus 1 of the present invention can analyze user psychological transitions related to the target keyword from a group of comment sentences transmitted for each contributor. The psychological transition analysis device 1 includes a psychological keyword dictionary unit 101, a score storage unit 102, a comment sentence acquisition unit 11, a psychological state determination unit 12, a score update unit 13, and an influencer detection unit 14 together with a communication interface unit. And have. These functional components excluding the communication interface unit are realized by executing a program that causes a computer installed in the apparatus to function. Moreover, the flow of processing of these functional components can be understood as a psychological transition analysis method of the apparatus.

[Psychological keyword dictionary unit 101]
The psychological keyword dictionary unit 101 registers psychological keywords in advance for each psychological state. The psychological state transitions in time series as follows, corresponding to the AIDMA model.
Cognitive stage: Attention: Attention->
Emotional stage: Interest->
Desire: Desire->
Memory: Memory->
Action stage: Action: Action

  The AIDMA model is an abbreviation that represents a consumer psychological process for advertising. By tracking utterances of the same user in a time series with respect to the target keyword (target product service), it can be estimated by applying to a transition model of purchasing psychological state in economics.

The psychological keyword dictionary unit 101 registers a psychological keyword for the target keyword (for example, “a company smartphone”) for each psychological state (A, ID, M, A).
Attention state: sale, CM, new work, new product, ...
Interest state: Lineup, brochure, HP, friends, ...
Desired state: Wanted, price, specs, design, ...
Memory state: shop, exhibition, real thing, preview, ...
Action state: buy, get, give up, model change This is a dictionary in which words characteristically appearing in each psychological state are registered for the target keyword.

The psychological keyword dictionary unit 101 is basically created manually in advance. However, pre-psychological keywords registered manually as a seed (teacher data), the characteristically appearing word to each psychological state, using statistical indicators (e.g. chi ² value and Akaike Information Criterion (AIC)) It may be calculated.

[Comment text acquisition unit 11]
The comment text acquisition unit 11 acquires the comment text from the SNS site server 2 for each user address (contributor) using the target keyword (for example, “a company smartphone”) as a search key in time series. The target keyword may be, for example, the name of a product service for which influencers are to be analyzed. This target keyword is set for a system user, for example, a promotion person in charge of a company. The acquired comment text is output to the psychological state determination unit 12.

  FIG. 3 is an explanatory diagram showing comment sentences transmitted in time series.

  According to FIG. 3, various comment texts of a large number of users are represented for the target keyword (a company smartphone). Here, the comment text includes a mixture of quoted addresses (user addresses of other users) quoted in its own comments.

According to FIG. 3, it is assumed that the psychological transition analysis apparatus 1 has analyzed the psychological state included in the comment text for the target keyword “a company smartphone”. For example, it is assumed that comment sentences are transmitted in the following time series.
User @CCC “When @ABC is a company smartphone release?”
User @ABC "New a company smartphone. Http: //news.html"
User @AAA “A company's new smartphone will be released?” @ABC: New company a smartphone
http: //news.html ””
User @AAA "Do not watch CM of a company's smartphone well."
User @BBB "I wonder if it will be changed to a company smartphone RT" @ABC: This is a new company a smartphone
http: //news.html ””
User @AAA “@DEF: Company a smartphone, the color changed when you looked at HP. Cute!”
User @DEF "New a company smartphone, easy to use and recommended"
User @AAA "I also want a company smartphone" @DEF: New company a smartphone, easy to use
recommendation""
・ ・ ・ 「・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・
User @AAA "Let's see a company smartphone at the shop this time"

  FIG. 4 is an explanatory diagram in which the comment text including the target keyword is divided for each user address.

According to FIG. 4, only the comment text transmitted from the user @AAA including the target keyword “a company smartphone” is shown.
User @AAA “A company's new smartphone will be released?” @ABC: New company a smartphone
http: //news.html ””
User @AAA "Do not watch CM of a company's smartphone well."
User @AAA “@DEF: Company a smartphone, the color changed when you looked at HP. Cute!”
User @AAA "I also want a company smartphone" @DEF: New company a smartphone, easy to use
recommendation""
User @AAA "Let's see a company smartphone at the shop this time"

[Psychological state determination unit 12]
If the psychological keyword of the psychological keyword dictionary unit 101 is included in the comment text for each contributor, the psychological state determination unit 12 outputs the psychological state of the psychological keyword. According to the present invention, for each of these user addresses, an effect contributing to the user's psychological transition is to be estimated.

  FIG. 5 is an explanatory diagram in which a psychological state is given to the comment text including the target keyword.

From the comment text including the target keyword “Company a” and the psychological keyword “new work”, it can be estimated that the user is in an attention state.
From the comment text including the target keyword “a company smartphone” and the psychological keyword “CM”, it can be estimated that the user is in an attention state.
From the comment text including the target keyword “a company smartphone” and the psychological keyword “HP”, it can be estimated that the user is in an interest state.
From the comment text including the target keyword “Company a smartphone” and the psychological keyword “I want”, it can be estimated that the user is in the Desire state.
From the comment text including the target keyword “a company smartphone” and the psychological keywords “shop” and “preview”, it can be estimated that the user is in a memory state.

  Here, when the psychological state of the user transitions between before and after browsing the comment text of the other user, it is determined that the comment text of the other user contributes to the transition of the psychological state. By accumulating the determination results, it is possible to estimate the influence level of the comment text of the other user.

As another embodiment, when a plurality of psychological keywords are included in one comment sentence, the psychological state of the entire comment sentence may be determined based on the appearance frequency. Alternatively, the psychological state may be determined using a discriminator such as a support vector machine from the appearance tendency of words in the teacher data. Further, the likelihood may be given in addition to the AIDMA model.
“A company's new smartphone will be released”
New work = A, release = A
A = 2, I = 0, D = 0, M = 0, A = 0
Likelihood of A = 100%

[Score storage unit 102]
The score storage unit 102 stores a score of each psychological state for each user address.

[Score Updater 13]
The score updating unit 13 has the following two steps.
<Psychological transition determination step>
<Score update step>

<Psychological transition determination step>
The score update unit 13 estimates a psychological transition in which psychological states are arranged in time series for each poster. For example, the following psychological transition is estimated.
A, A, I, A, I, I, D, I, D, D, M, M, M

Here, it is also preferable to assign a numerical value that increases in order of A->I->D->M-> A to the psychological transition. As a result, the psychological transition is obtained by arranging the numerical values of the psychological state in time series. It can be understood that the state has shifted to the next psychological state depending on the magnitude of the numerical value.
A = 0, I = 1, D = 2, M = 3
A, A, I, A, I, I, D, I, D, D, M, M, M
↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓
0,0,1,0,1,1,2,1,2,2,3,3,3
However, in this case, a psychological state may naturally occur.

(Correction by moving average of psychological transition)
On the other hand, the score update unit 13 corrects the numerical values arranged in time series so as to be in ascending order as much as possible. Specifically, the average value of the psychological states of adjacent n comments before and after is set as the psychological state of the comment. That is, it is also preferable to set the moving average value within a predetermined range in the previous stage.
Take the average of two adjacent terms.
A, A, I, A, I, I, D, I, D, D, M, M, M
↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓
0, 0, 1, 0, 1, 1, 2, 1, 2, 2, 3, 3, 3
0.3, 0.3, 0.4, 0.6, 1, 1, 1.4, 1.6, 2, 2.2, 2.4, 2.6, 2.75, 3
((((((((((((((Rounding rounded down ↓))))))))))))))))))
0, 0, 0, 1, 1,1,1, 2, 2,2, 2, 3, 3, 3
↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓
A, A, A, I, I, I, I, D, D, D, D, M, M, M

(Correction based on time difference between psychological state transitions)
The comment text includes the transmission time. Moreover, the psychological state determination part 12 outputs the transmission time of the comment sentence containing a psychological keyword with the said psychological state. Thus, the score updater 13 can register the elapsed time interval between the psychological states arranged in time series.

Then, the score update unit 13 weights the numerical values arranged in time series in the time series psychological transition as follows.
The longer the elapsed time interval in the previous stage, the smaller the weight for the value. The shorter the elapsed time interval in the previous stage, the greater the weight for the value.

As another embodiment, the product of the likelihood of the estimated psychological state may be used for numerical values of 0 to 4. When A = 10%, I = 70%, D = 20%, and M = 0%, multiplication is performed as follows.
0 × 10% + 1 × 70% + 2 × 20% + 3 × 0% = 1.1

<Score update step>
The score update unit 13 executes one or both of the following two update steps and instructs the score storage unit 102.
(Criteria update step) When a psychological state is changed between a reference comment sentence including a user address and one or more forward comment sentences in the past, the reference Increment the psychological score of the comment text.
(Backward update step) When the psychological state changes between the reference comment sentence and one or more backward comment sentences in the future, the score of the psychological state of the backward comment sentence is incremented.

Specifically, for each user, the AIDMA transition of the forward / backward comment text is expressed as follows, as seen from the standard comment text.
Forward comment text-> standard comment text: db (a, i, d, m, a ')
Standard comment text-> backward comment text: da (a, i, d, m, a ')
Each element of (a, i, d, m, a ′) is represented by 0/1.
From the reference comment text including the user address of the other user, it is considered that the influence degree of the other user's comment text increases as the user's psychological state approaches the purchase behavior between the preceding and following comments.

  FIG. 6 is an explanatory diagram showing the first score update in the present invention.

(S61) For user @ AAA, the psychological state is changing as follows.
“Do not watch CM of company a smartphone.”
(Foreign comment text)-> Attention A
"@DEF: Company a smartphone, the color changed when you looked at HP. It's cute!"
(Standard comment text)-> Interest I
Here, the user @AAA browses the comment text of the user @DEF, and the psychological state is changed from attention A-> interest I. At this time, it is determined that the comment text of user @DEF contributed to the transition of the psychological state to interest I. Therefore, the score value P (I) of the interest I in the user @DEF is incremented by 1 for the score storage unit 102.
Forward comment text-> standard comment text: db (0,1,0,0,0)

(S62) Next, the psychological state of user @AAA is changing as follows.
"@DEF: Company a smartphone, the color changed when you looked at HP. It's cute!"
(Standard comment text)-> Interest I
“I want a company smartphone too!” @DEF: A new company company smartphone, easy to use and recommended. ”
(Backward comment text)-> Desire D
Here, the user @AAA browses the comment text of @DEF, and the psychological state further changes from interest I to desire D. At this time, the comment text of @DEF is determined to have contributed to the transition of the psychological state to desire D. Therefore, the score value P (D) of desire in the user @DEF is incremented by 1 for the score storage unit 102.
Standard comment text-> backward comment text: da (0,0,1,0,0)

Finally, the score is estimated by accumulating db (a, i, d, m, a ′) and da (a, i, d, m, a ′) for each user address. For example, according to the score storage unit 102 of FIG.
User @ABC is recognized as an influencer having a relatively high contribution to the psychological states “attention (A)”, “interest (I)”, and “desire (D)”.
The user @DEF is recognized as an influencer having a relatively high contribution to the psychological states “memory (M)” and “behavior (A)”.

  FIG. 7 is an explanatory diagram showing the second score update in the present invention.

(S71) For user @ AAA, the psychological state is changing as follows.
"A company's new smartphone will be released?" @ABC: New company a smartphone
http: //news.html ””-> Attention A
“Do you often watch CM of company a smartphone?”-> Interest I
Here, the user @AAA browses the comment text of the user @ABC, so that the psychological state transitions to attention A-> interest I. At this time, it is determined that the comment text of the user @ABC contributed to the transition of the psychological state to the interest I. Therefore, the score value P (I) of the interest I in the user @ABC is incremented by 1 for the score storage unit 102.

(S72) Next, the psychological state of user @AAA is changing as follows.
“Do you often watch CM of company a smartphone?”-> Interest I
"@DEF: a company smartphone, a new product came out. It's cute!"-> Interest I
Here, even if the user @AAA browses the comment text of the user @DEF, the psychological state remains as interest I and does not change. Therefore, no score value is updated for the score storage unit 102.

(S73) Next, for user @ AAA, the psychological state is changing as follows.
"@DEF: a company smartphone, a new product came out. It's cute!"-> Interest I
"A company's new smartphone?" @DEF: The new company a smartphone is easy to use and recommended "
-> Interest I
Here, even if the user @AAA browses the comment text of the user @DEF, the psychological state remains as interest I and does not change. Therefore, no score value is updated for the score storage unit 102.

  FIG. 8 is an explanatory diagram showing the score update with the first weight in the present invention.

FIG. 8 shows the same score update as in FIG. However, as shown below, the score is incremented with a different weight for each psychological state.
(S81) Weight wb (Weight before) for the psychological state of the standard comment sentence
wb (A), wb (I), wb (D), wb (M), wb (A ')
(S82) Weight after the psychological state of the backward comment sentence wa (Weight after) (≠ wb)
wa (A), wa (I), wa (D), wa (M), wa (A ')
In this case, the influence degree P is expressed by the following equation using the state changes db (i) and da (i).
P = _{Σi = A to A ′} (wb (i) × db (i) + wa (i) × da (i))

According to FIG. 8, the weights wb (i) and wa (i) can be set in a 5 × 5 matrix. For example, b _ati means the weight for the transition from Attention to Interest. By adjusting the value of each element of this matrix, it is possible to calculate the degree of influence that emphasizes a specific state transition.

  The weight wb of the psychological state AIDM (Attention: attention-s> Interest: interest-> Desire: desire-> Memory: memory) is set larger than the weight wb of the psychological state A ′ (Action: action). When the score of the psychological state of the reference comment sentence is incremented, the user address described in the reference comment sentence shifted to the psychological state AIDM is compared with the user address described in the reference comment sentence shifted to the psychological state A ′. Means that the contribution to the psychological state is high.

  The weight wa of the psychological state AIDM (Attention: attention-s> Interest: interest-> Desire: desire-> Memory: memory) is set smaller than the weight wa of the psychological state A ′ (Action: action). When the score of the psychological state of the backward comment sentence is incremented, the user address of the user address described in the reference comment sentence transitioned to the psychological state AIDM is compared with the user address described in the reference comment sentence transitioned to the psychological state A ′. Means that the contribution to the psychological state is low.

  FIG. 9 is an explanatory diagram showing the score update with the second weight in the present invention.

According to FIG. 9, when the psychological state changes between the reference comment sentence and one or more backward comment sentences in the future, the score update unit 13 posts the time between the reference comment sentence and the backward comment sentence. The weight wa is set to be smaller as the interval becomes longer. That is, a plurality of comments are used in addition to one comment sentence adjacent to the reference comment sentence in front / back. In that case, the weight is set as follows.
A larger weight is set for a psychological state that has been transitioned by a comment sentence that is closer in time or number to the reference comment sentence.
A smaller weight is set for a psychological state that is shifted by a comment sentence that is distant from the reference comment sentence in terms of time or number of cases.

[Influencer detection unit 14]
The influencer detection unit 14 transmits a score for each psychological state stored in the score storage unit 102 for each user address. The score means the degree to which the comment text transmitted from the user affects the purchasing psychology of other users. For example, a user address whose score is equal to or greater than a predetermined threshold for the psychological state is detected as an influencer for the transition to the psychological state.

  As described above in detail, according to the program, apparatus, and method of the present invention, it is possible to detect an influencer having a high influence level of a comment text to be posted.

  That is, it is recognized that a user who sends a comment sentence that brings other users closer to purchasing behavior has a higher value as an influencer (a user who contributes to promotion). Instead of simply referring to the psychological state of the comment text, it is possible to estimate other users (influencers) that have an influence on the transition of the user's purchasing psychological state.

  Various changes, modifications, and omissions of the above-described various embodiments of the present invention can be easily made by those skilled in the art. The above description is merely an example, and is not intended to be restrictive. The invention is limited only as defined in the following claims and the equivalents thereto.

DESCRIPTION OF SYMBOLS 1 Psychological transition analyzer 101 Psychological keyword dictionary part 102 Score memory | storage part 11 Comment sentence acquisition part 12 Psychological state determination part 13 Score update part 14 Influencer detection part 2 SNS site server 3 Terminal

Claims (10)

A program that causes a computer installed in the device to function so as to analyze the effect of the user's psychological transition related to the target keyword from the group of comment sentences transmitted for each contributor,
Psychological keyword dictionary means for registering psychological keywords in advance for each psychological state;
For each contributor, comment sentence acquisition means for acquiring comment sentences including the target keyword in time series,
For each contributor, when the psychological keyword of the psychological keyword dictionary means is included in the comment text, the psychological state determining means for outputting the psychological state of the psychological keyword,
Score storage means for storing a score of each psychological state for each citation address;
In the case of a reference comment sentence including a quote address in a group of comment sentences, if the psychological state changes between the reference comment sentence and one or more forward comment sentences in the past, the psychological state of the reference comment sentence And / or if the psychological state changes between the reference comment sentence and one or more backward comment sentences in the future, the psychological state score of the backward comment sentence is incremented. Score update means for instructing the score storage means;
A program that causes a computer to function as an influencer detection means that outputs a predetermined number of cited addresses from the top of the score as influencers having a high influence level for each psychological state using the score storage means . Target keywords are keywords based on product services,
The program according to claim 1 or 2, further causing the computer to function such that the quoted address is a user address in SNS (Social Networking Service). Psychological keyword dictionary means that psychological state corresponds to AIDMA (Attention: attention-s> Interest: interest-> Desire: desire-> Memory: memory-> Action: behavior) model. The program according to claim 1 or 2, wherein the computer functions so as to register a psychological keyword for the keyword. The psychological state is assigned a numerical value that increases in the order of A->I->D->M-> A,
A time series of psychological transitions is a time series of psychological values.
The program according to claim 3, wherein the score updating unit causes the computer to function so that the numerical values arranged in time series are a moving average value within a predetermined range. The score updating means sets the weight wb for the psychological state of the reference comment sentence and the weight wa for the psychological state of the backward comment sentence as different values, and causes the computer to function so as to weight and increment the score. The program according to claim 3 or 4, characterized in that The score updating means causes the computer to function such that the weight wb for the psychological state of the reference comment sentence and the weight wa for the psychological state of the backward comment sentence are incremented with different weights for each psychological state. The program according to any one of claims 3 to 5. The weight wb of the psychological state AIDM (Attention: attention-s> Interest: interest-> Desire: desire-> Memory: memory) is larger than the weight wb of the psychological state A ′ (Action: action)
The weight wa of the psychological state AIDM (Attention: attention-s> Interest: interest-> Desire: desire-> Memory: memory) is set in advance so as to be smaller than the weight wa of the psychological state A ′ (Action: action). The program according to claim 5 or 6, wherein the computer functions as described above. The score update means includes
When the psychological state changes between the reference comment sentence and one or more backward comment sentences in the future, the weight wa becomes smaller as the posting time interval between the reference comment sentence and the backward comment sentence becomes longer. The program according to any one of claims 3 to 7, wherein the computer is caused to function so as to be set. A device that analyzes the effect of a user's psychological transition on a target keyword from a group of comment sentences sent to each contributor,
Psychological keyword dictionary means for registering psychological keywords in advance for each psychological state;
For each contributor, comment sentence acquisition means for acquiring comment sentences including the target keyword in time series,
For each contributor, when the psychological keyword of the psychological keyword dictionary means is included in the comment text, the psychological state determining means for outputting the psychological state of the psychological keyword,
Score storage means for storing a score of each psychological state for each citation address;
In the case of a reference comment sentence including a quote address in a group of comment sentences, if the psychological state changes between the reference comment sentence and one or more forward comment sentences in the past, the psychological state of the reference comment sentence And / or if the psychological state changes between the reference comment sentence and one or more backward comment sentences in the future, the psychological state score of the backward comment sentence is incremented. Score update means for instructing the score storage means;
An influencer detecting means for outputting a predetermined number of cited addresses from the top of the score as influencers having a high influence level for each psychological state using the score storage means. A method of analyzing the effect of a user's psychological transition on a target keyword from a group of comment sentences sent to each contributor using an apparatus,
The device is
For each psychological state, a psychological keyword dictionary part in which psychological keywords are registered in advance,
For each citation address, it has a score storage unit that stores the score of each psychological state,
The device is
For each contributor, a first step of acquiring comment sentences including the target keyword in time series;
For each contributor, if the comment text includes a psychological keyword in the psychological keyword dictionary, a second step of outputting the psychological state of the psychological keyword;
In the case of a reference comment sentence including a quote address in a group of comment sentences, if the psychological state changes between the reference comment sentence and one or more forward comment sentences in the past, the psychological state of the reference comment sentence And / or if the psychological state changes between the reference comment sentence and one or more backward comment sentences in the future, the psychological state score of the backward comment sentence is incremented. A third step of instructing the score storage means;
And a fourth step of outputting, for each psychological state, a predetermined number of cited addresses from the top of the score as influencers having a high influence level using the score storage means.

Images