JP6663945B2 - Program, apparatus, and method for estimating user's empathy for content - Google Patents

Description

  The present invention relates to a technique for estimating a user's empathy ability for content.

2. Description of the Related Art In recent years, an interactive system for interacting with a person has been installed in a smartphone application or a robot.
As in the past, there is a technology such as ELIZA or SHRDLU that responds to a user utterance in a question-and-answer format (for example, see Non-Patent Document 1). According to this technique, a question example close to a user's utterance is detected by pattern matching from a set of question examples stored in advance.
On the other hand, the current interactive system can be roughly classified into a “task-oriented interactive function” and a “chat interactive function”, and is configured by combining these functions.
There is also a dialog system technology that focuses on "social dialogue" for the purpose of establishing and maintaining interpersonal relationships (for example, see Non-Patent Document 2).
Further, there is a technique for enhancing the intimacy of a user with a dialogue system that applies empathy to social dialogue (for example, see Non-Patent Document 3).
Further, there is a technique in which an information recommendation function of a field of interest of a user is applied to a dialog system (for example, see Non-Patent Document 4).

Okumura Manabu, et al., "Dialogue System", Corona, 2017, ISBN: 978-4-339-02757-0 TW Bickmore and RW Picard, “Establishing and maintaining long-term human-computer relationships”, ACM Tran. On Computer-Human Interaction 'TOCHI), 12 (2): 293-327, 2005., [online], [Heisei 30 February 12, 2008], Internet <URL: http://www.ccs.neu.edu/home/bickmore/publications/toCHI.pdf> Ryuichiro Higashinaka, Koji Dosaka, Hideki Isozaki, "Effects of Empathy and Self-Disclosure in Dialogue Systems", The 15th Annual Meeting of the Linguistic Processing Society, 446-449, 2009., [online], [February 12, 2018 Date Search], Internet <URL: http://www.kecl.ntt.co.jp/icl/lirg/members/rh/pdf/NLP2009.pdf> Shiori Itoshima, Kazuto Tominaga, Hiroyuki Kameda, "Creating a bot that recommends information based on interest from dialogue," The 11th Forum on Information Technology, F-024, 2012., [online], [ February 12, 2018 Search], Internet <URL: https://ipsj.ixsq.nii.ac.jp/ej/?action=repository_action_common_download&item_id=151579&item_no=1&attribute_id=1&file_no=1> Koji Toda, "Psychometric Measures II: Capturing the Connection between Human and Society <Interpersonal Relationships and Values> 3. Empathy and Other Consciousness", edited by Tomio Yoshida, edited by Hiromichi Hori, Science, pp. 118-137 Yumi Suzuki, Kazuyo Kino, “Development of Multidimensional Empathy Scale (MES): Focusing on Self-Oriented / Other-Oriented Discrimination”, Educational Psychology Research, 56, 487-497, 2008. [online], [Searched February 12, 2018], Internet <URL: https://ci.nii.ac.jp/naid/110007028208> Item response theory, [online], [searched on February 12, 2018], Internet <URL: https://en.wikipedia.org/wiki/%E9%A0%85%E7%9B%AE%E5% BF% 9C% E7% AD% 94% E7% 90% 86% E8% AB% 96> Singular value decomposition, [online], [searched on February 12, 2018], Internet <https://en.wikipedia.org/wiki/%E7%89%B9%E7%95%B0%E5%80% A4% E5% 88% 86% E8% A7% A3>

According to the above-described related art, an attempt is made to increase the affinity of the user in the dialogue system, but no social dialogue based on knowledge of psychology has been studied.
According to the field of psychology, there is "empathy" as a property of recognizing other people's feelings and experiencing the same emotion as others (for example, see Non-Patent Document 5). However, empathy cannot be determined from only a single viewpoint. Specifically, empathy is a pluralistic expression of "the ability to accurately recognize another person's psychological state (recognition)" and "the strength of proxy response to another person's psychological state (emotional degree)". It is necessary to judge from an appropriate viewpoint (for example, see Non-Patent Document 6).

  For example, a user who browses an article understands the characters (including anthropomorphic ones) of the article and the thoughts and feelings of the author, and has some feelings. This is based on the user's empathy for the content. Users have different empathy abilities. For example, even for the same content, each user has different empathy.

  Therefore, an object of the present invention is to provide a program, an apparatus, and a program for estimating a user's empathy ability for content.

According to the present invention, there is provided a program for causing a computer mounted on a device for estimating a user's empathy ability θ _E of an estimation target user s _E to a content c to include:
As a learning function,
For each user _{s n (s 1 ~s N)} , empathy of each content _{c l (c 1 ~c L)} (t 11 ~t 1L, ~, t N1 ~t NL) empathy index matrix T that associates _{a sensitivity} matrix storage means for storing _nl ,
For each user s _n, and stores a feature value _{(f 11 ~f 1J, ~,} f N1 ~f NJ) user's feature quantity matrix F _nj that associates a given individual user characteristic element (v ₁ ~v _J) User feature matrix storage means;
For each user s _n, enter the sympathy of matrix T _nl of each content c _l, and item response theory calculating means for calculating the empathy theta _n based on item response theory IRT (Item Response Theory),
For each user s _n , function as a machine learning engine that learns as teacher data in which a user feature matrix F _nj and empathy ability θ _n are associated with each other,
As an operation function,
Estimation target user s _E, as a user characteristic amount matrix calculating means for calculating a feature value (f _E1 ~f _EJ) user's feature quantity matrix F _E which associates a given individual user characteristic element (v ₁ ~v _J) To make it work,
The machine learning engine is characterized by inputting a user feature matrix F _E of the estimation target user s _E and causing a computer to function so as to output an empathic ability θ _E.

According to another embodiment of the program of the present invention,
The sensitivity is
A degree of recognition t _R that understands the emotions of the characters or authors of the content,
An emotional level t _E that produces the same emotion as the character or author of the content,
It is also preferable that the item response theory calculating means causes the computer to function so as to calculate as a two-parameter logistic model.

According to another embodiment of the program of the present invention,
User characteristic element is based on a plurality of questions to the user s _n,
Feature amount, preferably also causes a computer to function as is based on the answer of the user s _n to the question.

According to another embodiment of the program of the present invention,
The content is posted on the Web server by an unspecified number of third parties,
It is also preferable to cause the computer to function such that the content influence M is based on the emotion received when the user browses the content on the Web server.

According to another embodiment of the program of the present invention,
The machine learning engine also preferably causes the computer to function as being based on linear regression, a support vector machine, or deep learning regression learning.

According to the present invention, there is provided an apparatus for estimating an empathic ability θ _E of an estimation target user s _E for a content c,
As a learning function,
For each user _{s n (s 1 ~s N)} , empathy of each content _{c l (c 1 ~c L)} (t 11 ~t 1L, ~, t N1 ~t NL) empathy index matrix T that associates _{a sensitivity} matrix storage means for storing _nl ,
For each user s _n, and stores a feature value _{(f 11 ~f 1J, ~,} f N1 ~f NJ) user's feature quantity matrix F _nj that associates a given individual user characteristic element (v ₁ ~v _J) User feature matrix storage means;
For each user s _n, enter the sympathy of matrix T _nl of each content c _l, and item response theory calculating means for calculating the empathy theta _n based on item response theory IRT (Item Response Theory),
A machine learning engine for learning as teacher data in which the user feature matrix F _nj and the empathy ability θ _n are associated with each other for each user s _n ;
As an operation function,
Estimation target user s _E, the user feature amount matrix calculating means for calculating a feature value (f _E1 ~f _EJ) user's feature quantity matrix F _E which associates a given individual user characteristic element (v ₁ ~v _J) Has,
Machine learning engine receives the user's feature quantity matrix F _E of the estimated target user s _E, and outputs the empathy theta _E.

According to the present invention, there is provided an estimation method for an apparatus for estimating a user's empathy ability θ _E of an estimation target user s _E for content c,
The equipment is
As a learning function,
For each user _{s n (s 1 ~s N)} , empathy of each content _{c l (c 1 ~c L)} (t 11 ~t 1L, ~, t N1 ~t NL) empathy index matrix T that associates _{a sensitivity} matrix storage unit for storing _nl ,
For each user s _n, and stores a feature value _{(f 11 ~f 1J, ~,} f N1 ~f NJ) user's feature quantity matrix F _nj that associates a given individual user characteristic element (v ₁ ~v _J) A user feature matrix storage unit,
For each user s _n, a eleventh step of inputting the sympathy of matrix T _nl of each content c _l, calculates the empathy theta _n based on item response theory IRT (Item Response Theory),
Using a machine learning engine, for each user s _n , executing a twelfth step of learning as teacher data in which the user feature amount matrix F _nj and the empathy ability θ _n are associated with each other;
As an operation function,
A twenty-first step of calculating a user feature matrix F _{E in} which feature amounts (f _{E1 to} f _EJ ) of predetermined user feature elements (v _{1 to} v _J ) are associated with the estimation target user s _E ;
Machine learning engine, enter the user's feature quantity matrix F _E of the estimated target user s _E, and executes a second 22 step of outputting the empathy theta _E.

  ADVANTAGE OF THE INVENTION According to the program, apparatus and method of this invention, the user's empathy ability with respect to a content can be estimated.

It is a system configuration diagram in the present invention. FIG. 3 is a functional configuration diagram of an estimation device according to the present invention. It is explanatory drawing showing a co-sensitivity matrix. FIG. 4 is an explanatory diagram illustrating a user feature amount matrix. FIG. 4 is an explanatory diagram showing a degree of recognition and an emotion of a user with respect to content. It is explanatory drawing which applied recognition degree and emotion degree as a co-sensitivity. It is explanatory drawing which applied recognition degree and emotion degree as a co-sensitivity matrix. FIG. 9 is an explanatory diagram showing an item response category characteristic curve for each content.

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

  FIG. 1 is a system configuration diagram in the present invention.

The Web server 2 connected to the Internet publishes various contents. The Web server 2 is, for example, a news site, a blog (Web log) site, a mini blog site (for example, twitter (registered trademark), an SNS (Social Networking Service) site (for example, facebook (registered trademark) or LINE (registered trademark)). The contents may be posted on a Web server by an unspecified number of third parties.
The terminal 3 accesses the Web server 2 via the access network and the Internet, and allows the user to browse those contents.
The estimating apparatus 1 according to the present invention estimates a user's empathy ability for content. That is, the user group can be classified according to psychological empathy.

  FIG. 2 is a functional configuration diagram of the estimation device according to the present invention.

The estimating apparatus 1 according to the present invention estimates the user's empathy ability θ _E of the estimation target user s _E for the content c.
According to FIG. 2, the estimation device 1 is roughly classified into a <learning function> and an <operation function>, and has a machine learning engine 12 in common for both functions.
The estimating apparatus 1 includes a content storage unit 101, a co-sensitivity matrix storage unit 102, a user feature amount matrix storage unit 103, and an item response theory calculation unit 11 as learning functions.
Further, it has a user feature amount matrix calculation unit 13 as an operation function.
These functional components are realized by executing a program that causes a computer mounted on the device to function. In addition, the flow of processing of these functional components can be understood as a method of estimating the device.

<Learning function>
Learning function, for each content c _l as a teacher data (c ₁ to c _L), a plurality of users s _n (s ₁ ~s _N) IRT result between (empathy theta _n), and user's feature quantity , A learning model is constructed in the machine learning engine 12.

[Content storage unit 101]
The content storage unit 101 stores a plurality of contents (c _{1 to} c _L ) as teacher data. The content may be in any format that can be viewed by the user, such as text (articles and blogs), images, videos, and the like.

[Sensitivity matrix storage unit 102]
Empathy index matrix storage unit 102, corresponding to each user (s ₁ ~s _N), empathy of each content _{c l (c 1 ~c L)} (t 11 ~t 1L, ~, t N1 ~t NL) to The attached sensitivity matrix T _nl is stored.

  FIG. 3 is an explanatory diagram showing a co-sensitivity matrix.

According to FIG. 3, for each content c _l as a teacher data, to obtain answers sympathy degree as subjective evaluation, for example, by questionnaire. For example, for each subject (s _{1 to} s _N ), each content (c _{1 to} c _L ) is specified, and for example, a three-level (w = 3) answer is obtained for the sympathy. The co-sensitivity matrix T _nl (t ₁₁ to t _1L ,..., T _N1 to t _NL ) thus created is stored.

[User feature matrix storage unit 103]
User characteristic quantity matrix storage unit 103, for each user s _n, associating feature value of each given user characteristic element _{(v 1 ~v J) (f} 11 ~f 1J, ~, f N1 ~f NJ) and The user feature matrix F _nj (N rows and J columns) is stored.

The user characteristic element may generally be profile information (gender, age, occupation, etc.) of the user. In particular, it is preferably based on the user's personality (recognition ability, physical ability, intellectual ability, personality, patience, etc.). The user characteristic element may be a parameter element that identifies the characteristics of the user according to the application.
The user characteristic element may be based on a plurality of questions to the user s _n. In that case, the user feature amount will be based on the answer of the user s _n to the question.

  FIG. 4 is an explanatory diagram illustrating a user feature amount matrix.

According to FIG. 4, each user as a subject answers a plurality of questions. The content of the question is preferably related to the user's empathy. Thus, answers matrix for each question qm each user s _n (N rows and M columns) is obtained.
In this answer matrix, each user's answer is compressed into J dimension by Singular Value Decomposition. The matrix U summarizing the compressed feature vector by all users s _n, and the user characteristic amount matrix.
The singular value decomposition is a matrix decomposition method for a matrix having a complex number or a real number as a component in linear algebra (for example, see Non-Patent Document 8). There is a decomposition of the matrix X of rank r.
^{X = UΣV T -> U =} X · R R = V · (V T · V) -1 · Σ -1
U: User feature matrix

[Item Response Theory Calculation Unit 11]
IRT calculation section 11, each user s _n, enter the sympathy of matrix T _nl of each content c _l, calculates the empathy theta _n based on item response theory IRT (Item Response Theory).

The item response theory refers to a test theory for measuring user ability θ (and discrimination degree α / difficulty degree β) based on a user's response to an evaluation item group (for example, see Non-Patent Document 7). In this method, the user ability (and the degree of discrimination / difficulty) is calculated stochastically from the user's discrete answers to the evaluation items. It is suitable for Computerized Adaptive Testing, which maintains a stock of test items for users and considers the difficulty of multiple tests to be equivalent.
User ability θ _l : real value representing the magnitude of the ability of user n with respect to content l (identification degree a _l : real value representing the ability of content l to identify the user's empathetic ability)
(Difficulty b _l : Real value indicating the difficulty of content l)
Basically, the difference between the user ability and the degree of discrimination and the degree of difficulty is applied to a logistic curve to determine the probability of an answer. For example, if the content c _l is high empathetic degree for users s _n, the probability is close to 1 as possible, when a low sympathetic degree Conversely, the probability is close to zero as possible.
In the case of a two-parameter logistic model, the probability is calculated as follows.
p (θ) = 1 / (1 + e− ^{Da (θ−b)} )

[Machine learning engine 12 for learning function]
The machine learning engine 12 of the learning function learns, for each user s _n , teacher data in which the user feature amount matrix F _nj is associated with the empathic ability θ _n .
The machine learning engine 12 may be based on linear regression, a support vector machine, or deep learning regression learning.

<Operation function>
Operational function, for the unknown user s _l, to estimate the empathy ability θ _l for the content.

[User feature matrix calculation unit 13]
The user feature matrix calculation unit 13 generates a user feature matrix F _{E in} which feature amounts (f _{E1 to} f _EJ ) of predetermined user feature elements (v _{1 to} v _J ) are associated with the estimation target user s _E. calculate.
Here, the user feature value matrix calculating unit 13, in exactly the same way as user characteristic quantity matrix storage unit 103 described above, calculates the user characteristic amount matrix F _E.

[Machine learning engine 12 of operation function]
Machine learning unit 12 of the operational function, enter a user's feature quantity matrix F _E estimation target user (unknown) s _E, and outputs the empathy theta _E.

  FIG. 5 is an explanatory diagram showing the degree of recognition and emotion of the user with respect to the content.

According to psychological findings, sympathy is represented by “cognition” and “emotionality”.
"Sympathy": The nature of a person who understands the emotions of others and has the same emotions as others "Awareness": The degree of understanding the emotions (psychological state) of others "Emotionality": The same emotions as others Degree of possession (strength of proxy response) The "other" here means a character or an author of the content.

According to FIG. 5, a large number of users (subjects) browse the same article, and each user asks "is it a topic that interests me?" It is the answer result as evaluation.
Interesting topics? : Do not pull = low recognition
: Pull = High recognition A topic that touches the heart? : Not echoing = low emotion
: Resonance = High emotional level According to the table in FIG. 5, there are users who feel "interesting" and "not resonate" for the same article, and "not interest". Some users feel that it "resonates with the heart."

FIG. 6 is an explanatory diagram in which the degree of cognition and the degree of emotion are applied.
FIG. 7 is an explanatory diagram in which the recognition level and the emotion level are applied as the co-sensitivity matrix.

According to FIG. 6, the co-sensitivity matrix storage unit 102 stores the cognition level T ¹ and the emotion level T ² separately as the co-sensitivity matrix T as compared with FIG.
According to FIG. 7, each content (c _{1 to} c _L ) is clearly indicated to each of N subjects in the group of subjects (s _{1 to} s _N ), and the degree of recognition and emotion is, for example, three levels (w = 3). Get the answer. Thus, ratings Attribute = 2 empathy of matrix _{^{T 1 nl (t 1 11 ~t}} 1 1L, ~, t 1 N1 ~t 1 NL) and _{^{T 2 nl (t 2 11 ~t}} 2 1L, ~, t ² _{N1 to} t ² _NL ) are input to the item response theory calculation unit 11.

On the other hand, the item response theory calculation unit 11 calculates the empathy ability θ for the user s _n (s _{1 to} s _N ) as a two-parameter logistic model as follows.
Awareness of empathy: θ ^{₁ 1} ~θ ¹ _N
Emotional degree of empathy: θ ² ₁ ~θ ² _N
If some users can answer randomly, the item response theory calculation unit 11 may calculate the three-parameter logistic model.

Machine learning engine 12, for each content c _l, a feature amount matrix F _l, the identification of alpha ¹ ₁ and difficulty beta ¹ ₁ for recognition, and identification of alpha ¹ ₂ and difficulty beta ¹ ₂ for emotion degree Is learned using teacher data associated with. Then, a learning model is constructed as follows.
Learning model for empathy ability θ ^{₁ 1} of awareness: M ¹
Learning model for empathy ability θ ¹ ₂ emotional level: M ²

Then, machine learning engine 12 inputs the feature quantity matrix F _E of the estimated target user s _E, and outputs the empathy theta ¹ _E for identification of, and empathy theta ² _E for emotional level.

  FIG. 8 is an explanatory diagram showing an item response category characteristic curve for each content.

The item response category characteristic curve (IRCCC) is created as follows.
Horizontal axis: Sympathetic ability value (left: low sensitivity, right: high sensitivity)
(Left: low discrimination, right: high discrimination)
(Left: low emotion, right: high emotion)
Vertical axis: probability p
Curve: each option ("do not sympathize""do not understand""empathize")
("Not interested", "don't know", "interest"
("Does not affect heart""Do not understand""Do not affect heart"

The item response category characteristic curve in FIG. 8 shows the co-sensitivity matrices T ¹ and T ^{2 of the} degree of recognition and emotion for ₃₀₀ users (s _{1 to} s ₃₀₀ ) for ₁₀₀ contents (c _{1 to} c ₁₀₀ ). , An eight-dimensional user feature matrix F.
The difficulty b is on the same scale as the empathy ability θ, and the discrimination a determines the slope of the item response category characteristic curve. The larger the slope of the curve and the larger the difference between the difficulty b and the empathic ability θ, the clearer the answer.

According to FIG. 8, the item response category characteristic curves of the co-sensitivity matrices of the degree of recognition and the degree of emotion are shown for the four contents (ID36, ID72, ID82, ID94). It can be seen that ID82 and ID94 have higher recognition than ID72. In addition, it can be seen that ID82 has a higher emotional level than ID94 and ID72.
As a result, it is understood that the ID 82 has high recognition and high emotion. In addition, it can be seen that ID94 and ID72 have low emotional degrees.
It should be noted that a high sympathy can also be obtained for the ID 72 estimated to have low sympathy for a certain content if the user's empathy ability is high.

  As described above in detail, according to the program, the estimation device, and the method of the present invention, it is possible to estimate a user's empathy ability for content.

  For the above-described various embodiments of the present invention, various changes, modifications, and omissions in the scope of the technical idea and viewpoint of the present invention can be easily performed by those skilled in the art. The foregoing description is merely an example, and is not intended to be limiting. The invention is limited only as defined by the following claims and equivalents thereof.

DESCRIPTION OF SYMBOLS 1 Estimation apparatus 101 Content storage unit 102 Co-sensitivity matrix storage unit 103 User feature matrix storage unit 11 Item response theory calculation unit 12 Machine learning engine 13 User feature matrix calculation unit 2 Web server 3 Terminal

Claims (7)

A program for causing a computer mounted on a device for estimating a user's empathy ability θ _E of the estimation target user s _{E with} respect to the content c,
As a learning function,
For each user _{s n (s 1 ~s N)} , empathy of each content _{c l (c 1 ~c L)} (t 11 ~t 1L, ~, t N1 ~t NL) empathy index matrix T that associates _{a sensitivity} matrix storage means for storing _nl ,
Wherein for each user s _n, the feature amount of each given user characteristic element _{(v 1 ~v J) (f} 11 ~f 1J, ~, f N1 ~f NJ) stores user characteristic amount matrix F _nj that associates User feature matrix storage means for performing
For each of the user s _n, the type the sympathy of matrix T _nl of each content c _l, and item response theory calculating means for calculating the empathy theta _n based on item response theory IRT (Item Response Theory),
For each user s _n , function as a machine learning engine that learns as teacher data in which the user feature amount matrix F _nj and the empathy ability θ _n are associated with each other;
As an operation function,
Estimation target user s _E, as a user characteristic amount matrix calculating means for calculating a feature value (f _E1 ~f _EJ) user's feature quantity matrix F _E which associates a given individual user characteristic element (v ₁ ~v _J) To make it work,
The program, wherein the machine learning engine inputs the user feature amount matrix F _E of the estimation target user s _E and causes a computer to function so as to output an empathic ability θ _E. The co-sensitivity is
A degree of recognition t _R that understands the emotions of the characters or authors of the content,
An emotional degree t _E that produces the same emotion as the character or author of the content,
The program according to claim 1, wherein the item response theory calculation unit causes a computer to function so as to calculate as a two-parameter logistic model. The user characteristic element is based on a plurality of questions to the user s _n,
The feature quantity, the program according to claim 1 or 2, characterized in that causes a computer to function as is based on the answer of the user s _n for the question. The content is posted on a Web server by an unspecified number of third parties,
4. The computer according to claim 1, wherein the content influence M causes the computer to function so as to be based on an emotion received when the user browses the content on a Web server. 5. Question answer program. The machine learning engine according to any one of claims 1 to 4, wherein the machine functions as a computer based on linear regression, support vector machine, or deep learning regression learning. The program described in section. An apparatus for estimating the empathic ability θ _E of the user of the estimation target user s _{E with} respect to the content c,
As a learning function,
For each user _{s n (s 1 ~s N)} , empathy of each content _{c l (c 1 ~c L)} (t 11 ~t 1L, ~, t N1 ~t NL) empathy index matrix T that associates _{a sensitivity} matrix storage means for storing _nl ,
Wherein for each user s _n, the feature amount of each given user characteristic element _{(v 1 ~v J) (f} 11 ~f 1J, ~, f N1 ~f NJ) stores user characteristic amount matrix F _nj that associates User feature matrix storage means for performing
For each of the user s _n, the type the sympathy of matrix T _nl of each content c _l, and item response theory calculating means for calculating the empathy theta _n based on item response theory IRT (Item Response Theory),
A machine learning engine that learns as teacher data in which the user feature matrix F _nj and the empathy ability θ _n are associated with each other for each user s _n ;
As an operation function,
Estimation target user s _E, the user feature amount matrix calculating means for calculating a feature value (f _E1 ~f _EJ) user's feature quantity matrix F _E which associates a given individual user characteristic element (v ₁ ~v _J) Has,
The machine learning engine receives the user's feature quantity matrix F _E of the estimated target user s _E, and outputs the empathy theta _E device. An estimation method of an apparatus for estimating a user's empathy ability θ _E of an estimation target user s _{E with} respect to a content c,
The device comprises:
As a learning function,
For each user _{s n (s 1 ~s N)} , empathy of each content _{c l (c 1 ~c L)} (t 11 ~t 1L, ~, t N1 ~t NL) empathy index matrix T that associates _{a sensitivity} matrix storage unit for storing _nl ,
Wherein for each user s _n, the feature amount of each given user characteristic element _{(v 1 ~v J) (f} 11 ~f 1J, ~, f N1 ~f NJ) stores user characteristic amount matrix F _nj that associates A user feature matrix storage unit
For each of the user s _n, a 11th step of the type the sympathy of matrix T _nl of each content c _l, calculates the empathy theta _n based on item response theory IRT (Item Response Theory),
Using a machine learning engine, for each of the users s _n , performing a twelfth step of learning as the teacher data in which the user feature amount matrix F _nj and the empathy ability θ _n are associated with each other,
As an operation function,
A twenty-first step of calculating a user feature matrix F _{E in} which feature amounts (f _{E1 to} f _EJ ) of predetermined user feature elements (v _{1 to} v _J ) are associated with the estimation target user s _E ;
The machine learning engine executes the 22nd step of inputting the user feature amount matrix F _E of the estimation target user s _E and outputting an empathic ability θ _E.

Images