JP2022060824A - Keyword extraction device, keyword extraction program, and speech generation device - Google Patents

Abstract

To provide a keyword extraction device, a keyword extraction program, and a speech generation device which can provide a user with "sympathy" and "new awareness of interest or taste" simultaneously.SOLUTION: A keyword extraction device 10 comprises: a user personality setting unit 11 which sets a personality type of a user; a robot interest level acquisition unit 131 which acquires interest levels about individual keywords corresponding to the personality type as robot interest levels; a user interest level acquisition unit 12 which acquires interest levels about the individual words of the user as user interest levels; and a keyword extraction unit 132 which calculates difference between the robot interest level and the user interest level about each of the individual keywords and extracts a plurality of kinds of keywords about which the difference is within a specific range, including keywords providing sympathy and keywords providing new awareness.SELECTED DRAWING: Figure 1

Description

The present invention relates to a technique for extracting keywords used by a robot for dialogue with humans.

Conventionally, technology has been studied in which a robot that watches TV or the like with a person speaks according to the contents of a program.
For example, in Patent Document 1, a robot that makes an utterance related to a program by extracting a keyword from the video, audio, and subtitle data of a television program and combining it with a template sentence stored in advance to generate an utterance sentence. Proposed.

Further, Patent Document 2 proposes a robot that recognizes a user's face and performs an operation in which the expression content is changed according to each user based on the user's information such as age and gender.
Non-Patent Document 1 proposes a device that divides a user's personality into four types and recommends a favorite image for each personality type based on Jung's psychology.
Patent Document 3 proposes a device that recognizes a user's preference and recommends content based on reaction information obtained by analyzing captured image data without the operation of a user who is watching television.

Japanese Unexamined Patent Publication No. 2019-185400 Japanese Unexamined Patent Publication No. 2013-99823 Japanese Unexamined Patent Publication No. 2012-9957

Marina Uemura et al., "Proposal of a method for recommending images according to the personality type of viewers", 2019 Winter Games of the Institute of Image Information and Television Engineers, 22B-2 (2019).

In a communication robot that interacts with humans, it is important for the robot to "sympathize" with the user by speaking and to "give new interests and preferences" to the user.

However, in the technique of Patent Document 1, since the robot extracts a keyword from the video or subtitle data of the TV program being viewed and the robot makes an utterance, the user happens to "aware of new interests and tastes" from the content of the utterance. However, it was difficult to properly extract keywords intended for such effects. Further, since the user's interests and preferences are not taken into consideration in the utterance, it is difficult for the robot to "sympathize" with the user's interests or preferences through the utterance.

Further, in the technologies of Patent Documents 2 and 3 and Non-Patent Document 1, since the robot operates according to the user's request, even if it is possible to "sympathize" with the feeling of the user's request, it corresponds to the user's request. It was difficult to provide the user with "awareness of new interests and tastes" because they do nothing but do.

An object of the present invention is to provide a keyword extraction device, a keyword extraction program, and an utterance generation device that can simultaneously provide a user with "empathy" and "awareness of new interests and tastes".

The keyword extraction device according to the present invention includes a user personality setting unit that sets a user's personality type, a robot interest degree acquisition unit that acquires the degree of interest for each keyword according to the personality type as a robot interest degree, and the user. The difference between the robot interest level and the user interest level is calculated for each keyword and the user interest level acquisition unit that acquires the interest level for each keyword as the user interest level, and the difference is within a specific range. It is provided with a keyword extraction unit that extracts a plurality of types of keywords including keywords that give sympathy and keywords that give new awareness.

The keyword extraction unit may extract the keyword whose difference is closest to 0 as the keyword giving the sympathy.

The keyword extraction unit may extract a keyword whose robot interest level is larger than that of the user interest level and whose difference is the largest as a keyword that gives a new awareness.

The keyword extraction unit may extract a keyword having a greater user interest level than the robot interest level and having the largest difference, as a keyword representing the individuality of the user.

The keyword extraction unit may extract keywords for which at least one of the robot interest level and the user interest level is predetermined or higher.

In the user interest degree acquisition unit, the same keyword is acquired from the program by the keyword acquisition unit for acquiring the keyword in the program, the line-of-sight direction determination unit for determining the line-of-sight direction of the user, and the keyword acquisition unit. It may be provided with an interest degree calculation unit for calculating the user interest degree for the keyword based on the ratio of the number of times that the line-of-sight direction is facing the screen of the program when the keyword is acquired. ..

The keyword extraction device may include an interest level updating unit that averages the user interest levels calculated at regular intervals and updates the user interest levels.

The keyword extraction device includes a user interest level output unit that outputs the user interest level to the management server, and the user interest level acquisition unit is the user interest level for each of a plurality of users set to the same personality type. You may acquire the degree of interest for each keyword according to the personality type, which is updated based on.

The robot system according to the present invention includes the keyword extraction device and an utterance generation device that generates utterance sentences by different logics based on the plurality of types of keywords extracted by the keyword extraction device.

The keyword extraction program according to the present invention is for operating a computer as the keyword extraction device.

According to the present invention, the robot can simultaneously provide the user with "empathy" and "awareness of new interests and tastes".

It is a figure which shows the functional structure of the robot system in an embodiment. It is a figure which illustrates the personality data in an embodiment. It is a figure which shows the calculation method of the user interest degree in an embodiment. It is a figure which shows the method of updating the interest data of a user in an embodiment. It is a flowchart illustrating the operation flow of the user interest degree acquisition part in an embodiment. It is a figure which illustrates the extraction method of the interest keyword for each kind in an embodiment. It is a figure which illustrates the data stored in the interest keyword DB in an embodiment. It is a flowchart which illustrates the operation flow of the interest keyword management part in an embodiment. It is a figure which illustrates the state of the update and acquisition of the personality data in the robot system in an embodiment. It is a figure which shows the update method of the personality data in an embodiment.

Hereinafter, an example of the embodiment of the present invention will be described.
The present embodiment provides a keyword extraction device that extracts keywords used for dialogue by a robot that is watching a TV program together with a person (user).

The keyword extraction device uses the user's personality data as the robot's interest data in order to make the keywords that the user is likely to be interested in as the robot's interest, while the frequency with which the user turns his or her eyes to the TV when watching a TV program. Is the user's interest data.
Then, the keyword extraction device compares the user's personality data (= robot's interest data) with the user's interest data, and determines the type of the robot's interest keyword from the difference in the degree of interest of each keyword. In addition to the "interest" keyword that allows the robot to sympathize with the robot through dialogue, the "awareness" keyword that gives the user awareness of new interests and the "individuality" keyword that expresses the user's individuality are extracted.

FIG. 1 is a diagram showing a functional configuration of the robot system 1 in the present embodiment.
The robot system 1 includes a keyword extraction device 10 and a speech generation device 20 incorporated in the robot, and a personality data management server 30, and the keyword extraction device 10 and the personality data management server 30 are communicated with each other.
A plurality of keyword extraction devices 10 may be connected to the personality data management server 30.

The keyword extraction device 10 manages interest data of one or a plurality of users, manages interest data of a robot according to a user's personality type, and extracts keywords for generating an utterance sentence.
The details of the keyword extraction process will be described later.

The utterance generation device 20 generates utterance sentences by different logics based on the empathy keyword, the notice keyword, and the individuality keyword extracted by the keyword extraction device 10.
It should be noted that a template prepared for each type of keyword, an existing method using a machine learning model, or the like can be applied to the generation of the utterance sentence.

The personality data management server 30 manages the degree of interest in a plurality of keywords as personality data by the personality database (DB) 31 for each personality type, and provides the keyword extraction device 10 in response to a request.
Further, the personality data management server 30 includes a personality data updating unit 32 that updates the personality DB 31. The update process by the personality data update unit 32 will be described later.

Here, the personality types are classified into four types, thinking, intuition, feelings, and sensations, which have different taste tendencies, based on Jung's psychology shown in Non-Patent Document 1 described above.

FIG. 2 is a diagram illustrating personality data in the present embodiment.
In the personality data, the general interest level of a person of this personality type is set in the range of 0 to 1.0 for each keyword for each of the four personality types.

Personality data is a type of person classified according to general preference trends. People's tastes change significantly in a short period of time due to being influenced by television, etc., but personality data for each personality type rarely changes significantly in a short period of time. Therefore, in the keyword extraction method of the present embodiment, the degree of interest of the keyword for each personality type is set as the interest data of the robot, and the characteristic keyword is extracted by comparing with the interest data of the user who fluctuates greatly. ..

The keyword extraction device 10 is an information processing device provided with various interfaces in addition to a control unit and a storage unit, and the control unit executes software (keyword extraction program) stored in the storage unit to execute the present embodiment. Various functions are realized.

The control unit of the keyword extraction device 10 includes a user personality setting unit 11, a user interest level acquisition unit 12, an interest keyword management unit 13, a keyword output unit 14, and an interest data output unit 15.
Further, the storage unit of the keyword extraction device 10 includes an interest degree database (DB) 16 and an interest keyword database (DB) 17.

The user personality setting unit 11 sets the personality type of the user. The personality type can be determined by performing a personality determination test based on Jung's psychology.
In the present embodiment, for example, the user performs a personality determination test in advance, the determination result is input to the user personality setting unit 11, and is provided to the interest keyword management unit 13.

The user interest level acquisition unit 12 is a function unit that acquires the user interest level, which is the degree of interest of each keyword of the user, and includes the keyword acquisition unit 121, the line-of-sight direction determination unit 122, the viewing count analysis unit 123, and the interest level. It includes a calculation unit 124 and an interest level update unit 125.

The keyword acquisition unit 121 sequentially acquires keywords such as people, landscapes, and objects detected by using existing techniques such as object detection and prominence estimation from the video of a television program. In addition, keywords may be extracted from audio, subtitles, etc. synchronized with the video.

The line-of-sight direction determination unit 122 determines the user's line-of-sight direction from the image of the user when watching a television program acquired from a camera provided in the robot.
Specifically, the line-of-sight direction determination unit 122 detects the user's line of sight from the camera image of the user's face, and determines whether or not the line of sight is on the television screen. For example, a small line-of-sight detection sensor and existing image processing technology are used.

In the viewing count analysis unit 123, when the video of the keyword or the keyword itself appears in the TV program based on the keyword acquired by the keyword acquisition unit 121 and the line-of-sight direction of the user when the keyword is acquired, the user can use the television. Accumulate the number of times the line of sight is directed in the direction.
Here, the same scene in which the keyword appears, or the case where the line of sight is directed more than once in a predetermined period including the time when the keyword appears is counted as one time. For example, if you are not interested and accidentally turn your gaze to the TV screen, it is unlikely that your gaze will turn to the same keyword again, so the higher the degree of interest, the greater the number of times.

The interest level calculation unit 124 calculates the user interest level of each keyword by using the number of times the same keyword appears in the TV program viewed and the number of times the user turns his / her eyes to the screen on which this keyword appears.

FIG. 3 is a diagram showing a method of calculating a user interest level in the present embodiment.
For example, when the keyword "cityscape" is acquired from a TV program, the number of times the cityscape appears in the watched program l = 10, and the number of times the user looks at the TV screen when the cityscape appears m = Let it be 5. At this time, the user's interest level s for the user's keyword "cityscape" is
s = m / l = 5/10 = 0.5
Is calculated.
As a result, a combination of the keyword including "cityscape (0.5)" and the user's interest level can be obtained as the user's interest data.

The interest level update unit 125 periodically updates the user's interest level stored in the interest level DB 16, that is, the user's interest level for each keyword, for example, once a day.

FIG. 4 is a diagram showing a method of updating user interest data in the present embodiment.
For example, it is assumed that the user interest level for the cityscape is calculated as a = 0.8 by the interest level calculation unit 124 for a certain period (for example, 24 hours).
At this time, assuming that the user interest level stored in the interest level DB 16 is s = 0.6, the user interest level x after the update is the average value so far using the update count n (= 1) so far. As a result, it is calculated as follows.
x = s + (as) / (n + 1) = 0.6 + (0.8-0.6) / 2 = 0.7

Here, as the number of updates n increases, the fluctuation of the user interest level x decreases. Therefore, an upper limit may be set for the value so that the number of updates n does not become too large. As a result, the user interest level that reflects the recent tendency of the user is calculated.
For example, when the current user interest level s = 0.2 and the new user interest level a = 0.8, when the number of updates n = 99, the updated user interest level is x = 0.2 + (0.8). -0.2) /100=0.206, which is a slight change, but when the upper limit of n = 4 is set, x = 0.2 + (0.8-0.2) /5=0.32. , The change of interest is reflected in the user's interest level.

FIG. 5 is a flowchart illustrating the operation flow of the user interest level acquisition unit 12 in the present embodiment.
This process is executed for a certain period of time such as 24 hours, and each parameter is initialized and executed repeatedly.

In step S1, the keyword acquisition unit 121 acquires a keyword from a television program and adds 1 to the number of occurrences l of this keyword.
In step S2, the line-of-sight direction determination unit 122 analyzes the line-of-sight direction of the user and determines whether or not the user has turned to the television screen within the period in which the keyword appears. If this determination is YES, the process proceeds to step S3, and if the determination is NO, the process proceeds to step S4.

In step S3, the viewing count analysis unit 123 adds 1 to the viewing count m for the acquired keyword.
In step S4, the user interest level acquisition unit 12 determines whether or not a certain period of time has elapsed from the start and the timing of aggregation has come. If this determination is YES, the process proceeds to step S5, and if the determination is NO, the process returns to step S1.

In step S5, the interest level calculation unit 124 calculates the user interest level for each keyword within the period based on the number of appearances l of the keyword and the number of views m of the user for this keyword.
In step S6, the interest level updating unit 125 updates the user's interest data stored in the interest level DB 16 based on the calculated user interest level.

The interest keyword management unit 13 is a functional unit that extracts and manages interest keywords for each type of "empathy", "awareness", and "individuality", and includes a robot interest degree acquisition unit 131 and a keyword extraction unit 132.

When the robot interest level acquisition unit 131 acquires the user's personality type from the user personality setting unit 11, it extracts the personality data corresponding to the corresponding personality type from the personality DB 31 of the personality data management server 30, and uses it as the robot's interest data. Memorize the combination of keywords and interest level (robot interest level).

The keyword extraction unit 132 acquires the user's interest data from the interest degree DB 16 and calculates the difference between the robot interest degree and the user interest degree for each keyword. Then, the keyword extraction unit extracts the sympathy keyword, the noticed keyword, and the individuality keyword in which the calculated difference is in a specific range, and stores them in the interest keyword DB 17.
The keyword extraction unit 132 may update the interest keyword DB 17 at the same frequency as the interest update unit 125 updates the interest DB 16, for example, once a day.

FIG. 6 is a diagram illustrating a method of extracting an interest keyword for each type in the present embodiment.
For example, when the robot user sets the "sensory" type as the personality type, the "sensory" type personality data is copied to the robot's interest data.

When the user's interest data is created depending on the user's program viewing status, the keyword extraction unit 132 compares the robot's interest data with the user's interest data and calculates the difference in the degree of interest for each of the two keywords. ..

When the degree of interest in the robot is high and the degree of interest in the user is also high (the difference in the degree of interest is small), it is defined as an empathy keyword because the degree of interest in both the user and the robot is high.
I noticed that keywords with high robot interest and low user interest (large difference in interest) are keywords that are generally of interest due to the nature of the user, and may be of interest in the future. Defined as a keyword.
Keywords with low robot interest and high user interest are defined as individual keywords.

Therefore, the keyword extraction unit 132 extracts the keyword whose difference (= robot interest degree-user interest degree) is closest to 0 as an empathy keyword and stores it in the interest keyword DB 17.
Further, the keyword extraction unit 132 extracts the keyword whose robot interest level is larger than the user interest level and whose difference is the largest as the noticed keyword, and stores it in the interest keyword DB 17.
Further, the keyword extraction unit 132 extracts a keyword having a user interest level higher than the robot interest level and having the largest difference as an individual keyword and stores it in the interest keyword DB 17.

The keyword extraction unit 132 may extract these types of interest keywords from keywords in which at least one of the robot interest degree and the user interest degree is predetermined or higher.
In addition, a predetermined number of keywords may be extracted for each type of empathy, awareness, and individuality.

FIG. 7 is a diagram illustrating data stored in the interest keyword DB 17 in the present embodiment.
The interest keyword DB 17 stores extracted keywords for each type of empathy, awareness, and individuality. Further, the robot interest level, the user interest level, and the difference at the time of extraction for these keywords may be stored in association with each other.

FIG. 8 is a flowchart illustrating the flow of operation of the interest keyword management unit 13 in the present embodiment.
In step S11, the robot interest degree acquisition unit 131 acquires the personality type of the user set by the user personality setting unit 11.
In step S12, the robot interest degree acquisition unit 131 acquires personality data corresponding to the user's personality type from the personality DB 31.
In step S13, the robot interest degree acquisition unit 131 sets the acquired personality data as the robot interest data.

In step S14, the keyword extraction unit 132 acquires the user's interest data from the interest level DB 16.
In step S15, the keyword extraction unit 132 calculates the difference between the robot interest level and the user interest level.
In step S16, the keyword extraction unit 132 extracts the interest keyword based on the calculated difference and stores it in the interest keyword DB 17.

The keyword output unit 14 transmits the keyword stored in the interest keyword DB 17 to the utterance generation device 20 in response to a request from the utterance generation device 20 or when the data of the interest keyword DB 17 changes. Then, let the robot speak for dialogue with the user.

The interest data output unit 15 outputs the user's interest data stored in the interest level DB 16 to the personality data management server 30 together with the user's personality type.
The personality data update unit 32 of the personality data management server 30 receives interest data for each user from the keyword extraction devices 10 provided for each of the plurality of robots, and updates the personality DB 31 for each personality type.

FIG. 9 is a diagram illustrating a state of updating and acquiring personality data in the robot system 1 in the present embodiment.
The interest data A for each user acquired by the keyword extraction device 10 is integrated with the personality type B for each user, and is appropriately transmitted to the personality data management server 30 as data C for updating the personality data.

At this time, the number of users of the robot may be multiple. When a plurality of users have the same personality type, a plurality of records of interest data may be provided for each personality type, or one record with an average degree of interest may be provided.

Similarly, in the other robots, the data for updating the personality data is transmitted to the personality data management server 30.
The personality data management server 30 updates the degree of interest of each keyword for each personality type according to a predetermined procedure based on the received data for updating the plurality of personality data.
The updated personality data is appropriately acquired from the keyword extraction device 10.

FIG. 10 is a diagram showing a method of updating personality data in the present embodiment.
In the personality data stored in the personality DB 31, the degree of interest in a certain keyword of a certain personality type is c, and the degree of user interest in the same keyword of a user of the same personality type acquired from the keyword extraction device 10 is x. Using the number of updates as n, the degree of interest y after the update is calculated as an average value so far as follows.
y = c + (x−c) / (n + 1)

For example, if the thinking type user has an interest level of 0.6 for the keyword "sea", the interest level 0.8 for the thinking type "sea" in the personality DB 31 is assumed to be the number of updates n = 1. ,
0.8 + (0.6-0.8) /2=0.7
Will be updated to.
The personality data update unit 32 may set an upper limit on the number of updates n, as in the interest level update unit 125.

According to the present embodiment, the keyword extraction device 10 calculates a difference between the robot interest level according to the personality type of the user and the user interest level, and based on the value of this difference, the empathy keyword, the noticed keyword, and the individuality. Extract each keyword.
As a result, the keyword extraction device 10 can simultaneously provide the user with "empathy" and "awareness of new interests and tastes" by causing the robot to speak with the extracted keywords.

Specifically, the keyword extraction device 10 extracts the keyword having the closest difference to 0 as an empathy keyword, and notices the keyword whose robot interest level is larger than the user interest level and whose difference is the largest. The keywords that have a higher user interest level than the robot interest level and the largest difference are extracted as individual keywords.
As a result, the keyword extraction unit can easily extract the most characteristic interest keywords based on the difference in the degree of interest.

When the robot interest level is low and the user interest level is high, it is a user's unique keyword that is different from the tendency according to the personality type. Therefore, the utterance generator 20 generates an utterance sentence based on this keyword. It can give users an opportunity to talk about their interests. That is, it is possible to satisfy the user's desire to speak by uttering a word that encourages the user to talk about a matter that the user is familiar with but is not well known to others.

At this time, the robot can realize a natural dialogue with suitable utterance sentences according to the type of the keyword of interest by generating utterance sentences with a plurality of types of keywords extracted by the keyword extraction device 10 by different logics. ..
Further, the keyword extraction device 10 can suppress the extraction of keywords that are unlikely to be of interest by extracting keywords in which at least one of the robot interest level and the user interest level is predetermined or higher.

The keyword extraction device 10 acquires the keyword from the program and determines the direction of the user's line of sight, thereby determining the user's interest level based on the ratio of the number of times the keyword appears in the program to the screen. calculate.
As a result, the keyword extraction device 10 can calculate an appropriate degree of interest in the keyword based on the actual viewing behavior of the user.

The keyword extraction device 10 calculates the user interest level at regular intervals, and updates the user interest level by averaging the values calculated in the past.
As a result, the keyword extraction device 10 can appropriately reflect the tendency so far in the user's interest data without being greatly affected by the temporary fluctuation of interest.

When the keyword extraction device 10 transmits the user interest level to the personality data management server 30, the personality data management server 30 can collect interest data of a plurality of users and update appropriate personality data for each personality type. Further, the keyword extraction device 10 can extract an appropriate keyword by using the latest personality data.

As a result, the personality data is shared among the plurality of robots, and the personality data can be updated. The personality data management server 30 collects interest data of many users who use the robot and keeps updating the personality data, so that the number of keywords included in the personality data increases.
As a result, in extracting the three types of interest keywords, the keyword extraction device 10 can reduce the possibility that the keyword in the user's interest data is not in the personality data and the difference in interest degree cannot be calculated. Further, since the personality DB 31 is updated based on the interest data of many users, the number of samples is large, so that the reliability of the personality data is improved.

Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments. Moreover, the effects described in the above-described embodiments are merely a list of the most preferable effects resulting from the present invention, and the effects according to the present invention are not limited to those described in the embodiments.

In the above-described embodiment, the keyword extraction device 10 has been described as being incorporated in the robot, but the present invention is not limited to this, and the keyword extraction device 10 is arranged outside the robot and is connected to the robot by wire or wirelessly or via a network. May be done.

In the above-described embodiment, the personality data corresponding to the personality type of the user is copied as it is as the interest data of the robot, but the present invention is not limited to this. The interest data of the robot may be obtained by multiplying the interest level of the personality data by a predetermined weight.
For example, by multiplying the personality data by a coefficient> 1, a higher degree of interest in the keywords than the general interest according to the personality type is set in the robot's interest data. This makes it easier for each keyword to be classified as "empathy" or "awareness" rather than "individuality."
In this way, the keyword extraction device 10 can give individuality to the robot, such as a robot that sympathizes well and a robot that gives new awareness.

In the above-described embodiment, the personality data management server 30 holds the personality DB 31, but the present invention is not limited to this. The personality DB 31 may be configured such that the robot stores it in advance and operates by the robot alone without communicating with an external server.

In the present embodiment, the configuration and operation of the keyword extraction device 10 have been mainly described, but the present invention is not limited to this, and each component may be provided and configured as a method or a program for extracting keywords. ..

Further, even if a program for realizing the function of the keyword extraction device 10 is recorded on a computer-readable recording medium, the program recorded on the recording medium is read into a computer system and executed. good.

The term "computer system" as used herein includes hardware such as an OS and peripheral devices. Further, the "computer-readable recording medium" refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, or a CD-ROM, and a storage device such as a hard disk built in a computer system.

Furthermore, a "computer-readable recording medium" is a communication line that dynamically holds a program for a short period of time, such as a communication line when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. It may also include a program that holds a program for a certain period of time, such as a volatile memory inside a computer system that is a server or a client in that case. Further, the above program may be for realizing a part of the above-mentioned functions, and may be further realized for realizing the above-mentioned functions in combination with a program already recorded in the computer system. ..

1 Robot system 10 Keyword extraction device 11 User personality setting unit 12 User interest level acquisition unit 13 Interest keyword management unit 14 Keyword output unit 15 Interest data output unit 16 Interest level DB
17 Interesting keyword DB
20 Utterance generator 30 Personality data management server 31 Personality DB
32 Personality data update unit 121 Keyword acquisition unit 122 Line-of-sight direction determination unit 123 Views analysis unit 124 Interest level calculation unit 125 Interest level update unit 131 Robot interest level acquisition unit 132 Keyword extraction unit

Claims (10)

The user personality setting unit that sets the user's personality type,
A robot interest level acquisition unit that acquires the degree of interest for each keyword according to the personality type as a robot interest level, and a robot interest level acquisition unit.
A user interest level acquisition unit that acquires the user interest level for each keyword as a user interest level,
For each keyword, a keyword that calculates the difference between the robot interest level and the user interest level, and extracts a plurality of types of keywords including a keyword that gives empathy and a keyword that gives new awareness, in which the difference is within a specific range. A keyword extraction device equipped with an extraction unit. The keyword extraction device according to claim 1, wherein the keyword extraction unit extracts the keyword whose difference is closest to 0 as the keyword that gives the sympathy. The keyword extraction unit according to claim 1 or 2, wherein the keyword extraction unit extracts a keyword whose robot interest level is larger than that of the user interest level and whose difference is the largest as a keyword that gives a new awareness. Keyword extractor. Any of claims 1 to 3, wherein the keyword extraction unit extracts a keyword having a user interest level higher than that of the robot interest level and having a maximum difference size as a keyword representing the individuality of the user. Keyword extractor described in Crab. The keyword extraction device according to any one of claims 1 to 4, wherein the keyword extraction unit extracts keywords in which at least one of the robot interest degree and the user interest degree is a predetermined value or more. The user interest acquisition unit
The keyword acquisition section that acquires keywords in the program, and
The line-of-sight direction determination unit that determines the line-of-sight direction of the user,
The user for the keyword is based on the ratio of the line-of-sight direction facing the screen of the program when the keyword is acquired, out of the number of times the same keyword is acquired from the program by the keyword acquisition unit. The keyword extraction device according to any one of claims 1 to 5, further comprising an interest level calculation unit for calculating the degree of interest. The keyword extraction device according to claim 6, further comprising an interest level updating unit that averages the user interest levels calculated at regular intervals and updates the user interest levels. A user interest level output unit that outputs the user interest level to the management server is provided.
From claim 1, the user interest acquisition unit acquires the interest level for each keyword according to the personality type, which is updated based on the user interest level for each of a plurality of users set to the same personality type. The keyword extraction device according to any one of claims 7. The keyword extraction device according to any one of claims 1 to 8,
A robot system including an utterance generation device that generates utterance sentences by different logics based on the plurality of types of keywords extracted by the keyword extraction device. A keyword extraction program for operating a computer as the keyword extraction device according to any one of claims 1 to 8.

Images