JP2022038498A - Selection program, selection method and selection device - Google Patents

Abstract

To select participants with high facilitator ability as facilitators.SOLUTION: A selection device acquires voice information including voices of a plurality of speakers. The selection device detects utterance information in which an utterance section in which an utterance included in the voice information is made and a speaker who made the utterance in the utterance section are associated with each other. The selection device performs voice recognition on the voice information and extracts words included in the voice information. The selection device evaluates impressions of the plurality of speakers before and after the utterance information including specific words included in the voice information. The selection device selects a facilitator based on the evaluation of the impressions of the plurality of speakers.SELECTED DRAWING: Figure 1

Description

The present invention relates to a selection device and the like.

In recent years, a system that supports exchanges at events such as conferences and exchange meetings has been known. Such a system supports exchanges by grouping participants with common hobbies and selecting facilitators.

For example, there is a technique for selecting an event secretary from candidates for secretary by candidacy or recommendation. By using this technology, it is possible to prepare for the event even if the secretary has not been decided.

Japanese Unexamined Patent Publication No. 2018-124750 Japanese Unexamined Patent Publication No. 2019-8130 Japanese Unexamined Patent Publication No. 2019-61129 International Publication No. 2017/168663

However, with the above-mentioned technique, it may not be possible to select a participant with high facilitator ability as a facilitator.

For example, in an event that includes a large number of people who meet for the first time, when the grouped participants are allowed to interact with each other, the success of the group exchange depends largely on how the facilitator is selected. The elected facilitators will partition the venue according to the agenda, but if an inexperienced person recommends it, the venue for interaction will not be lively. With the above-mentioned technology, although the willingness to interact can be understood from the number of times of participation in the event, it is not possible to judge whether the selected person is good at listening to the story, so it may not be possible to select a facilitator that excites the interaction.

In one aspect, it is an object of the present invention to provide a selection program, a selection method, and a selection device for selecting a participant having a high facilitator ability as a facilitator.

In the first plan, the computer is made to perform the following processing. The computer acquires voice information including the voices of a plurality of speakers. The computer detects the utterance information in which the utterance section in which the utterance is made and the speaker who made the utterance in the utterance section are associated with each other. The computer performs voice recognition on the voice information and extracts words included in the voice information. The computer evaluates the impressions of a plurality of speakers before and after the utterance information including a specific word contained in the voice information. The computer selects a facilitator based on the evaluation of the impressions of multiple speakers.

According to one embodiment, participants with high facilitator ability can be selected as facilitators.

FIG. 1 is a diagram for explaining an example of processing of the selection apparatus according to the first embodiment. FIG. 2 is a diagram showing an example of the system according to the first embodiment. FIG. 3 is a functional block diagram showing the configuration of the selection device according to the first embodiment. FIG. 4 is a diagram showing an example of a data structure of utterance information. FIG. 5 is a diagram showing an example of a data structure of utterance impression evaluation information. FIG. 6 is a diagram showing an example of a data structure of parrot return generation information. FIG. 7 is a diagram showing an example of a data structure of facilitating force evaluation information. FIG. 8 is a diagram showing an example of a data structure of participant rating information. FIG. 9 is a flowchart showing a processing procedure of the selection apparatus according to the first embodiment. FIG. 10 is a subroutine showing a processing procedure for detecting utterance information. FIG. 11 is a subroutine showing a processing procedure for specifying the parrot return. FIG. 12 is a functional block diagram showing the configuration of the selection device according to the second embodiment. FIG. 13 is a flowchart showing a processing procedure of the selection apparatus according to the second embodiment. FIG. 14 is a diagram showing an example of a hardware configuration of a computer that realizes the same function as the selection device.

Hereinafter, examples of the selection program, selection method, and selection device disclosed in the present application will be described with reference to the drawings. The present invention is not limited to this embodiment. In addition, each embodiment can be appropriately combined within a consistent range.

FIG. 1 is a diagram for explaining an example of processing of the selection apparatus according to the first embodiment. The selection device according to the first embodiment acquires voice information including voices of a plurality of speakers who are participants in a conference or an exchange meeting. Then, the selection device detects the utterance information in which the utterance section in which the utterance is performed and the utterance speaker in this utterance section, which are included in the acquired voice information, are associated with each other. The utterance information shown in FIG. 1 includes an utterance ID corresponding to the utterance section, a speaker corresponding to the utterance ID, and a start time and end time of the utterance. Further, the selection device associates the character string generated by performing voice recognition with the voice information with the utterance ID and stores it as the utterance content character string.

Further, the selection device extracts words included in this utterance content character string. Then, the selection device identifies parrot returns in which the extracted words match and the speakers are different in the adjacent utterance information in chronological order. Echolalia means that the speaker speaks the word spoken by the previous speaker as it is. Specifically, in the utterance information of FIG. 1, the word "commercial" W1 included in the utterance ID "h11" and the word "commercial" W2 included in the utterance ID "h21" are extracted, and these utterances are made. The parrot return is specified by different persons.

In addition, the selection device evaluates the participants' impressions of the voice information. The utterance impression evaluation information shown in FIG. 1 includes an utterance impression ID that identifies the utterance impression evaluation information, a utterance ID corresponding to this utterance impression ID, a time corresponding to the impression evaluation value, and an evaluation and quantification of the participant's impression. The impression evaluation value that was made is included.

Subsequently, the selection device determines whether or not the impression is improved by returning the parrot. Specifically, on the time axis shown in FIG. 1, it is determined whether or not the impression evaluation value V1 of the utterance ID “h12” corresponding to the utterance section immediately after the utterance ID “h21” in which the parrot is returned has increased. do.

After that, the selection device rates the participants by adding points to the participants whose impression evaluation value is increased by returning the parrot, and selects the facilitator using the rating result.

As described above, the selection device according to the first embodiment selects the participants who have improved the impression of the participants by returning the parrots as facilitators. As a result, participants with high facilitator ability can be selected as facilitators.

Next, the configuration of the system according to the first embodiment will be described. FIG. 2 is a diagram showing an example of the system according to the first embodiment. As shown in FIG. 2, this system has a microphone terminal 10 and a selection device 100. For example, the microphone terminal 10 and the selection device 100 are wirelessly connected to each other. The microphone terminal 10 and the selection device 100 may be connected by wire.

The microphone terminal 10 is a device for recording voice. The microphone terminal 10 transmits voice information to the selection device 100. The voice information includes voice information of speakers A to E who are participants in a conference or an exchange meeting. The microphone terminal 10 may include a plurality of microphones. When the microphone terminal 10 includes a plurality of microphones, the voice information collected by each microphone is transmitted to the selection device 100.

The selection device 100 acquires voice information from the microphone terminal 10, and selects, among the speakers A to E, the participants who are determined to have high facilitating ability as facilitators.

FIG. 3 is a functional block diagram showing the configuration of the selection device according to the first embodiment. As shown in FIG. 3, the selection device 100 includes a communication unit 110, an input unit 120, a display unit 130, a storage unit 140, and a control unit 150.

The communication unit 110 is a processing unit that wirelessly executes data communication with the microphone terminal 10. The communication unit 110 is an example of a communication device. The communication unit 110 receives voice information from the microphone terminal 10 and outputs the received voice information to the control unit 150. The selection device 100 may be connected to the microphone terminal 10 by wire. The selection device 100 may be connected to a network by the communication unit 110 to transmit / receive data to / from an external device (not shown).

The input unit 120 is an input device for inputting various information to the selection device 100. The input unit 120 corresponds to a keyboard, a mouse, a touch panel, and the like.

The display unit 130 is a display device that displays information output from the control unit 150. The display unit 130 corresponds to a liquid crystal display, a touch panel, or the like.

The storage unit 140 includes a voice buffer 140a, learning acoustic feature information 140b, utterance information 140c, utterance impression evaluation information 140d, parrot return generation information 140e, facilitating ability evaluation information 140f, and participant rating information 140g. Have. The storage unit 140 corresponds to a semiconductor memory element such as a RAM (Random Access Memory) and a flash memory (Flash Memory), and a storage device such as an HDD (Hard Disk Drive).

The voice buffer 140a is a buffer for storing voice information transmitted from the microphone terminal 10. In the voice information, the voice signal and the time are associated with each other.

The learning acoustic feature information 140b is information on the acoustic features of the voices of the speakers A to E that are learned in advance. Acoustic features include pitch frequency, frame power, formant frequency, and voice arrival direction. For example, the learning acoustic feature information 140b is a vector having values of pitch frequency, frame power, formant frequency, and voice arrival direction as elements.

The utterance information 140c is information in which the utterance section in which the utterance is made and the utterance speaker in this utterance section, which are included in the voice information of the participants, are associated with each other. FIG. 4 is a diagram showing an example of a data structure of utterance information. The utterance information 140c shown in FIG. 4 includes an utterance ID corresponding to the utterance section, a speaker corresponding to the utterance ID, and a start time and end time of the utterance. The utterance information includes a utterance content character string generated by performing voice recognition on the voice information.

The utterance impression evaluation information 140d is information that evaluates the impression of the participants. FIG. 5 is a diagram showing an example of a data structure of utterance impression evaluation information. The utterance impression evaluation information 140d shown in FIG. 5 includes an utterance impression ID that identifies each utterance impression evaluation information, a utterance section (speech ID) corresponding to this utterance impression ID, a time corresponding to the impression evaluation value, and a participant's impression. The impression evaluation value that is evaluated and quantified is included.

The parrot return generation information 140e is information indicating the occurrence of parrot return in which the extracted words match and the speakers are different in the adjacent utterance information in chronological order. FIG. 6 is a diagram showing an example of a data structure of parrot return generation information. The parrot return generation information 140e shown in FIG. 6 includes an parrot return ID that identifies each parrot return generation information, an utterance section (utterance ID) corresponding to this parrot return generation ID, and a speaker corresponding to this utterance ID. ..

The facilitating ability evaluation information 140f is information for evaluating the facilitating ability of the participants. FIG. 7 is a diagram showing an example of a data structure of facilitating force evaluation information. The facilitating power evaluation information 140f shown in FIG. 7 includes an evaluation information for each speaker, each speaker, and an evaluation value for each speaker. The evaluation information is generated every time the parrot return occurs, and a value of 1 is given when the impression is improved by the parrot return, 0 is given when the impression does not change, and -1 is given when the impression is bad. .. The evaluation value is the average value of the evaluation information, and is calculated by dividing the total of the evaluation information by the number of occurrences of parrot return.

Participant rating information 140g is information obtained by rating participants. FIG. 8 is a diagram showing an example of a data structure of participant rating information. The participant rating information 140g shown in FIG. 8 includes each speaker, participation information representing the history in which each speaker has acquired the highest evaluation value, and a rating for each speaker. The participation information indicates that, for example, the speaker E has obtained the highest evaluation value of 0.7 in the event (E3) in which the speaker E participated for the third time. The rating is added by 1 according to the number of times the highest value is recorded in the participation information.

The control unit 150 includes an acquisition unit 150a, an utterance information detection unit 150b, a voice recognition unit 150c, an utterance impression evaluation unit 150d, a specific unit 150e, a determination unit 150f, and a selection unit 150g. The control unit 150 is realized by hard-wired logic such as a CPU (Central Processing Unit), an MPU (Micro Processing Unit), an ASIC (Application Specific Integrated Circuit), and an FPGA (Field Programmable Gate Array).

The acquisition unit 150a is a processing unit that acquires voice information from the microphone terminal 10 via the communication unit 110. The acquisition unit 150a sequentially stores the voice information in the voice buffer 140a.

The utterance information detection unit 150b is a processing unit that acquires voice information from the voice buffer 140a and detects the utterance information 140c shown in FIG. 4 from the voice information. The utterance information detection unit 150b performs utterance section detection processing, acoustic analysis processing, and similarity evaluation processing.

First, an example of the "utterance section detection process" executed by the utterance information detection unit 150b will be described. The utterance information detection unit 150b identifies the power of the voice information, and detects a section sandwiched between silent sections whose power is less than the threshold value as the utterance section. The utterance information detection unit 150b may detect the utterance section by using the technique disclosed in International Publication No. 2009/145192.

The utterance information detection unit 150b divides the voice information divided by the utterance section into fixed-length frames. The utterance information detection unit 150b sets a frame number for identifying the frame of each frame. The utterance information detection unit 150b executes acoustic analysis processing and similarity evaluation processing, which will be described later, for each frame.

Subsequently, an example of the "acoustic analysis process" executed by the utterance information detection unit 150b will be described. For example, the utterance information detection unit 150b calculates the acoustic characteristics based on each frame of the utterance section included in the voice information. The utterance information detection unit 150b calculates the pitch frequency, the frame power, the formant frequency, and the voice arrival direction as acoustic features.

An example of a process in which the utterance information detection unit 150b calculates the “pitch frequency” as an acoustic feature will be described. The utterance information detection unit 150b calculates the pitch frequency p (n) of the audio signal included in the frame by using the estimation method of RAPT (A Robust Algorithm for Pitch Tracking). "N" indicates a frame number. The speech information detection unit 150b is "D.Talkin," A Robust Algorithm for Pitch Tracking (RAPT), "in Speech Coding & Synthesis, WB Kleijn and KK Pailwal (Eds.), Elsevier, pp.495-518, 1995". The pitch frequency may be calculated using the technique described in.

An example of a process in which the utterance information detection unit 150b calculates "frame power" as an acoustic feature will be described. For example, the utterance information detection unit 150b calculates the power S (n) in a frame having a predetermined length based on the equation (1). In the formula (1), "n" indicates a frame number, "M" indicates a time length of one frame (for example, 20 ms), and "t" indicates a time. “C (t)” indicates an audio signal at time t. The utterance information detection unit 150b may calculate the time-smoothed power as the frame power using a predetermined smoothing coefficient.

An example of a process in which the utterance information detection unit 150b calculates the “formant frequency” as an acoustic feature will be described. The utterance information detection unit 150b performs linear prediction (Linear Prediction Coding) analysis on the voice signal C (t) included in the frame, and calculates a plurality of formant frequencies by extracting a plurality of peaks. For example, the utterance information detection unit 150b calculates the first formant frequency: F1, the second formant frequency: F2, and the third formant frequency: F3 in ascending order of frequency. The utterance information detection unit 150b may calculate the formant frequency by using the technique disclosed in Japanese Patent Application Laid-Open No. 62-54297.

An example of a process in which the utterance information detection unit 150b calculates the “voice arrival direction” as an acoustic feature will be described. The utterance information detection unit 150b calculates the voice arrival direction based on the phase difference of the voice information recorded in the two microphones.

In this case, the utterance information detection unit 150b detects the utterance section from each voice information recorded by the plurality of microphones of the microphone terminal 10, compares the voice information of the frames of the same time in each utterance section, and ranks. Calculate the phase difference. The utterance information detection unit 150b may calculate the voice arrival direction by using the technique disclosed in Japanese Patent Application Laid-Open No. 2008-175733.

The utterance information detection unit 150b calculates the acoustic characteristics of each frame included in the utterance section of the voice information by executing the above acoustic analysis process. The utterance information detection unit 150b may use at least one of the pitch frequency, the frame power, the formant frequency, and the voice arrival direction as the acoustic feature, or may use a plurality of combinations as the acoustic feature. good. In the following description, the acoustic feature of each frame included in the utterance section of the voice information is referred to as "evaluation target acoustic feature".

Subsequently, an example of the "similarity evaluation process" executed by the utterance information detection unit 150b will be described. The utterance information detection unit 150b calculates the degree of similarity between the evaluation target acoustic feature of each frame of the utterance section and the learning acoustic feature information 140b.

For example, the utterance information detection unit 150b may calculate the Pearson product-moment correlation coefficient as the similarity, or may calculate the similarity using the Euclidean distance.

A case where the utterance information detection unit 150b calculates the Pearson product-moment correlation coefficient as the degree of similarity will be described. Pearson's product-moment correlation coefficient cor is calculated by Eq. (2). In the equation (2), "X" is a vector whose elements are the pitch frequency, frame power, formant frequency, and voice arrival direction of each of the acoustic features of the speakers A to E included in the learning acoustic feature information 140b. be. “Y” is a vector whose elements are the pitch frequency, frame power, formant frequency, and voice arrival direction of the acoustic feature to be evaluated. “I” is a number indicating an element of the vector. The utterance information detection unit 150b specifies a frame of the evaluation target acoustic feature having a Pearson product-moment correlation coefficient cor of the threshold value Thc or more as a frame including any of the voices of the speakers A to E. For example, the threshold Thc is set to "0.7". The threshold Thc may be changed as appropriate.

A case where the utterance information detection unit 150b calculates the similarity using the Euclidean distance will be described. The Euclidean distance d is calculated by the equation (3), and the similarity R is calculated by the equation (4). In the equation ( ₃ ), a1 to _ai correspond to the pitch frequency, frame power, formant frequency, and voice arrival direction value of each of the acoustic features of the speakers A to E included in the learning acoustic feature information 140b. b ₁ to bi correspond to the pitch frequency, frame power, _formant frequency, and voice arrival direction value of the acoustic feature to be evaluated. The utterance information detection unit 150b specifies a frame of the evaluation target acoustic feature whose similarity R is equal to or higher than the threshold value Thr as a frame including any of the voices of the speakers A to E. For example, the threshold Thr is set to "0.7". The threshold Thr may be changed as appropriate.

R = 1 / (1 + d) ... (4)

The utterance information detection unit 150b specifies a frame of the evaluation target acoustic feature whose similarity is equal to or higher than the threshold value as a frame including any of the voices of the speakers A to E. In other words, the utterance information detection unit 150b identifies the utterances A to E for each frame from the voice information.

The utterance information detection unit 150b repeatedly executes the above process to identify the speaker for all utterance sections. The utterance information detection unit 150b includes the start time and end time of each utterance section in the utterance information, and stores the utterance information 140c in the storage unit 140.

The voice recognition unit 150c is a processing unit that acquires voice information, performs voice recognition on the voice information, and generates an utterance content character string shown in FIG. The voice recognition unit 150c associates the character string generated by voice recognition with each utterance section, includes it in the utterance information 140c as an utterance content character string, and stores it in the storage unit 140. Further, the voice recognition unit 150c is a processing unit that extracts a word from the generated character string. The voice recognition unit 150c extracts words included in the utterance content character string.

The voice recognition unit 150c may use any technique to convert the voice information into a character string. For example, the voice recognition unit 150c converts voice information into a character string by using the technique disclosed in Japanese Patent Application Laid-Open No. 4-255900.

The utterance impression evaluation unit 150d is a processing unit that acquires voice information and evaluates the participant's impression of the voice information based on the vertical width of the pitch frequency of the voice signal in the voice information. The utterance impression evaluation unit 150d performs the "utterance impression evaluation process".

An example of the "utterance impression evaluation process" executed by the utterance impression evaluation unit 150d will be described. The utterance impression evaluation unit 150d acquires the audio signal of each utterance section and calculates the vertical width of the pitch frequency for each frame. Then, the speech impression evaluation unit 150d determines that the larger the vertical width of the pitch frequency of the audio signal is, the better the impression is, with the case where the impression is normal as the reference value 0, and gives an impression evaluation value having a large positive absolute value. However, it is determined that the smaller the vertical width of the pitch frequency of the audio signal is, the worse the impression is, and the impression evaluation value having a large negative absolute value is given. Then, the utterance impression evaluation unit 150d stores the impression evaluation value in the storage unit 140 as the utterance impression evaluation information 140d in association with the time. In addition, the utterance impression evaluation unit 150d may evaluate the participant's impression using biological information such as the participant's pulse. The utterance impression evaluation unit 150d may use the average value of the impressions of the speakers A to E as the impression evaluation value.

The specific unit 150e is a processing unit that acquires the utterance information 140c and specifies the parrot return from the utterance content character string. The specific unit 150e executes the parrot return specific process.

An example of the "echolalia specific process" executed by the specific unit 150e will be described. The specific unit 150e acquires the utterance information 140c, and identifies the utterance information in which the words included in the utterance content character string match and the utterance speakers are different in the adjacent utterance information in chronological order as the parrot return. Then, the specific unit 150e stores the specified parrot return as the parrot return generation information 140e in the storage unit 140. The utterance information adjacent to each other in chronological order refers to utterance information adjacent to each other in chronological order, but may be adjacent utterance information via one or more utterance information. That is, the case where the same word is spoken immediately after a certain utterance information may be specified as Echolalia, but the case where the same word is spoken with another person's utterance in between the utterance information may be specified as Echolalia. It may be included in the return.

The determination unit 150f is a processing unit that acquires the utterance impression evaluation information 140d and the parrot return generation information 140e, and determines whether or not the impression is improved by the parrot return. The determination unit 150f executes the determination process.

An example of the "determination process" executed by the determination unit 150f will be described. The determination unit 150f identifies an utterance section in which one parrot return occurs selected from the acquired parrot return generation information 140e, and in an utterance section in which the speaker immediately after the parrot return selected in the acquired speech impression evaluation information 140d is different. , Judge whether the impression evaluation value is increasing. Then, the determination unit 150f determines that the impression has improved if the impression evaluation value has increased, determines that the impression has not changed if the impression evaluation value has not changed, and determines that the impression has not changed if the impression evaluation value has decreased. It is judged that the impression has deteriorated. The determination unit 150f calculates the evaluation value by averaging the evaluation information, and stores the determination result in the storage unit 140 as the evaluation information and the evaluation value shown in FIG. 7.

The selection unit 150g is a processing unit that acquires facilitator force evaluation information 140f and selects a facilitator based on the judgment result of the determination unit 150f. The selection unit 150g executes the selection process.

An example of the "selection process" executed by the selection unit 150g will be described. Based on the acquired facilitating ability evaluation information 140f, the selection unit 150g updates the participant rating information 140g so as to raise the rating of the speaker with the highest evaluation value by 1, and stores it in the storage unit 140. Then, the selection unit 150g selects the speaker with the highest rating as a facilitator.

Next, an example of the processing procedure of the selection device 100 according to the first embodiment will be described. FIG. 9 is a flowchart showing a processing procedure of the selection apparatus according to the first embodiment. As shown in FIG. 9, the selection device 100 acquires the voice data of the speakers A to E who are participants of the exchange meeting or the like as a preliminary preparation, analyzes the acquired acoustic data, and determines the acoustic characteristics of each speaker. Calculate (step S101). For this advance preparation, voice data of an exchange meeting or the like held in the past may be used, or voice data acquired when the speakers A to E are registered in the selection device 100 may be used, or exchange may be performed. The voice data acquired at the time of self-introduction or chat at the beginning of a meeting or the like may be used.

Subsequently, the acquisition unit 150a of the selection device 100 acquires voice information including the voices of the plurality of speakers A to E and stores the voice information in the voice buffer 140a (step S102).

After that, the utterance information detection unit 150b of the selection device 100 detects the utterance information 140c from the voice information (step S103). FIG. 10 is a subroutine showing a processing procedure for detecting utterance information. As shown in FIG. 10, the utterance information detection unit 150b detects the utterance section from the acquired voice information (step S1301). Subsequently, the utterance information detection unit 150b calculates the acoustic characteristics for each frame included in each utterance section (step S1302). Further, the utterance information detection unit 150b calculates the similarity between the calculated evaluation target acoustic feature and the learning acoustic feature calculated in step S101, and identifies the speaker (step S1303). Then, the utterance information detection unit 150b stores the utterance information 140c in which the utterance section and the speaker are associated with each other in the storage unit 140 (step S1304). After that, the utterance information detection unit 150b determines whether or not the utterance section is detected from all the voice information (step S1305). When the utterance information detection unit 150b determines that the utterance section has not been detected from all the voice information (step S1305: No), the process returns to step S1301 and the process is repeated. On the other hand, when it is determined that the utterance information detection unit 150b has detected the utterance section from all the voice information (step S1305: Yes). Exit this subroutine.

Returning to FIG. 9, the voice recognition unit 150c of the selection device 100 performs voice recognition on the voice information and extracts a word (step S104).

Further, the utterance impression evaluation unit 150d of the selection device 100 evaluates the participant's impression of the voice information (step S105).

Subsequently, the specifying unit 150e of the selection device 100 specifies the parrot return (step S106). FIG. 11 is a subroutine showing a processing procedure for specifying the parrot return. As shown in FIG. 11, the specific unit 150e sets the speaker of the first utterance information in the time series as the last speaker (step S1601). Subsequently, the specific unit 150e determines whether or not the speaker of the next utterance information matches the final speaker in chronological order (step S1602).

When the specific unit 150e determines that the speaker of the next utterance information matches the final speaker in chronological order (step S1602: Yes), the process returns to step S1601. On the other hand, when the specific unit 150e determines that the speaker of the next utterance information does not match the final speaker in the time series (step S1602: No), the specific unit 150e is the speaker of the next utterance information in the time series. Is set as the last speaker (step S1603).

After that, the specific unit 150e determines whether or not the words match in the utterance information set by the final speaker and the immediately preceding utterance information (step S1604). When the specific unit 150e determines that the words match (step S1604: Yes), the specific unit 150e identifies the occurrence of parrot return and stores the parrot return generation information 140e in the storage unit 140 (step S1605). On the other hand, when the specific unit 150e determines that the words do not match (step S1604: No), the process proceeds to step S1606.

In step S1606, the specific unit 150e determines whether or not the utterance section of the last speaker is the last in the time series. When the specific unit 150e determines that the utterance section of the last speaker is not the last in the time series (step S1606: No), the process returns to step S1602 and the process is repeated. On the other hand, when the specific unit 150e determines that the utterance section of the last speaker is the last in the time series (step S1606: Yes), this subroutine is terminated.

Returning to FIG. 9, the determination unit 150f of the selection device 100 determines whether or not the impression of the speakers A to E is improved by returning the parrot (step S107).

Subsequently, the selection unit 150g of the selection device 100 evaluates the participants based on the determination result of the determination unit 150f, and stores the participant rating information 140g in the storage unit 140 (step S108). Then, the selection unit 150g selects the speaker with the highest rating as the facilitator based on the participant rating information 140g (step S109), and a series of processes is completed.

Next, the effect of the selection device 100 according to the first embodiment will be described. The selection device 100 identifies the parrot return and selects the participant who has improved the impression of the participant by the parrot return as the facilitator. Here, it is important that high facilitation ability means high ability to listen to others, and it is considered that people with high ability to listen to others often use Echolalia in conversation. .. Therefore, by selecting participants who improve the impression of participants by returning parrots as facilitators, participants with high facilitating ability can be selected as facilitators, and it is possible to excite exchanges and the like.

The selection device 100 not only identifies the parrot return, but also determines whether or not the impression has improved and selects the facilitator. This makes it possible to select participants with high facilitating ability as facilitators with higher accuracy than when specifying only parrot return.

Next, the detection device according to the second embodiment will be described. It is assumed that the system according to the second embodiment is wirelessly connected to the microphone terminal 10 in the same manner as the system described with reference to FIG. 3 of the first embodiment. Also in the second embodiment, the microphone terminal 10 records the voices of the speakers A to E and outputs the voice information.

The selection device according to the second embodiment acquires voice information from the microphone terminal 10, and selects the participants A to E who are determined to have high facilitating ability as facilitators.

FIG. 12 is a functional block diagram showing the configuration of the selection device according to the second embodiment. As shown in FIG. 12, the selection device 200 includes a communication unit 210, an input unit 220, a display unit 230, a storage unit 240, and a control unit 250.

The communication unit 210 is a processing unit that wirelessly executes data communication with the microphone terminal 10. The communication unit 210 is an example of a communication device. The communication unit 210 receives voice information from the microphone terminal 10 and outputs the received voice information to the control unit 250. The selection device 200 may be connected to the microphone terminal 10 by wire. The selection device 200 may be connected to the network by the communication unit 210 to transmit / receive data to / from an external device (not shown).

The input unit 220 is an input device for inputting various information to the selection device 200. The input unit 220 corresponds to a keyboard, a mouse, a touch panel, and the like.

The display unit 230 is a display device that displays information output from the control unit 250. The display unit 230 corresponds to a liquid crystal display, a touch panel, or the like.

The storage unit 240 includes a voice buffer 240a, learning acoustic feature information 240b, utterance information 240c, utterance impression evaluation information 240d, parrot return generation information 240e, facilitating power evaluation information 240f, and participant rating information 240g. , Complimentary word specific information 240h. The storage unit 240 corresponds to a semiconductor memory element such as a RAM or a flash memory, or a storage device such as an HDD.

The voice buffer 240a is a buffer for storing voice information transmitted from the microphone terminal 10. In the voice information, the voice signal and the time are associated with each other.

The learning acoustic feature information 240b is information on the acoustic features of the voices of the speakers A to E that are learned in advance. Acoustic features include pitch frequency, frame power, formant frequency, and voice arrival direction. For example, the learning acoustic feature information 240b is a vector having values of pitch frequency, frame power, formant frequency, and voice arrival direction as elements.

The utterance information 240c is information in which the utterance section in which the utterance is made and the utterance speaker in this utterance section, which are included in the voice information of the participants, are associated with each other.

The utterance impression evaluation information 240d is information that evaluates the participant's impression of the voice information.

The parrot return generation information 240e is information indicating the occurrence of parrot return in which the extracted words match and the speakers are different.

The facilitating ability evaluation information 240f is information for evaluating the facilitating ability of the participants.

Participant rating information 240g is information obtained by rating participants.

The praise word specific information 240h is information indicating that the praise word registered in advance has been specified. The compliment word specific information 240h is information stored in association with the specified compliment word and the time when the compliment word is spoken.

The control unit 250 includes an acquisition unit 250a, an utterance information detection unit 250b, a voice recognition unit 250c, an utterance impression evaluation unit 250d, a specific unit 250e, a determination unit 250f, a selection unit 250g, and a compliment specific unit 250h. And have. The control unit 250 is realized by a hard-wired logic such as a CPU, an MPU, an ASIC, or an FPGA.

The acquisition unit 250a is a processing unit that acquires voice information from the microphone terminal 10 via the communication unit 210. The acquisition unit 250a sequentially stores the voice information in the voice buffer 240a.

The utterance information detection unit 250b is a processing unit that acquires voice information from the voice buffer 240a and detects the utterance information 240c shown in FIG. 4 from the voice information. The utterance information detection unit 250b performs utterance section detection processing, acoustic analysis processing, and similarity evaluation processing.

The utterance section detection process, the acoustic analysis process, and the similarity evaluation process executed by the utterance information detection unit 250b are the same as those of the utterance impression evaluation unit 150d described in the first embodiment.

The voice recognition unit 250c is a processing unit that acquires voice information, performs voice recognition on the voice information, and generates an utterance content character string shown in FIG. The voice recognition unit 250c associates the character string generated by voice recognition with each utterance section, includes it in the utterance information 240c as an utterance content character string, and stores it in the storage unit 240. Further, the voice recognition unit 250c is a processing unit that extracts a word from the generated character string. The voice recognition unit 250c extracts words included in the utterance content character string.

The utterance impression evaluation unit 250d is a processing unit that acquires voice information and evaluates the participant's impression of the voice information based on the vertical width of the pitch frequency of the voice signal in the voice information. The utterance impression evaluation unit 250d performs the "utterance impression evaluation process".

The utterance impression evaluation process executed by the utterance impression evaluation unit 250d is the same as that of the utterance impression evaluation unit 150d described in the first embodiment.

The specific unit 250e is a processing unit that acquires the utterance information 240c and specifies the parrot return from the utterance content character string. The specific unit 250e executes the parrot return specific process.

The parrot return specifying process executed by the specific unit 250e is the same as that of the specific unit 150e described in the first embodiment.

The determination unit 250f is a processing unit that acquires the utterance impression evaluation information 240d, the parrot return generation information 240e, and the praise word specific information, and determines whether or not the impression is improved by the parrot return. The determination unit 250f executes the determination process.

An example of the "determination process" executed by the determination unit 250f will be described. The determination unit 250f identifies an utterance section in which one parrot return occurs selected from the acquired parrot return generation information 240e, and in an utterance section in which the speaker immediately after the parrot return selected in the acquired speech impression evaluation information 240d is different. , It is judged that the impression is improved when the impression evaluation value is increased or the compliment is specified. Further, the determination unit 250f identifies the utterance section in which one parrot return occurs selected from the acquired parrot return generation information 240e, and the speaker immediately after the parrot return selected in the acquired speech impression evaluation information 240d is a different utterance. It may be determined that the impression is improved when the impression evaluation value is increased and the compliment is specified in the section.

The selection unit 250g is a processing unit that acquires facilitator force evaluation information 240f and selects a facilitator based on the determination result of the determination unit 250f. The selection unit 250g executes the selection process.

The selection process executed by the selection unit 250g is the same as that of the selection unit 150g described in the first embodiment.

The praise word specifying unit 250h is a processing unit that acquires utterance information 240c and identifies praise words from words to others. The compliment word specifying unit 250h executes the compliment word specifying process.

An example of the "praise word identification process" executed by the compliment word identification unit 250h will be described. The praise word specifying unit 250h acquires the utterance information 240c and identifies a word matching the pre-registered praise word from the words included in the utterance content character string. Complimentary words are words such as "wow". Then, the praise word specifying unit 250h stores the specified praise word and the time when the praise word is uttered in the storage unit 240 as the praise word specific information 240h in association with each other.

Next, an example of the processing procedure of the selection device 200 according to the second embodiment will be described. FIG. 13 is a flowchart showing a processing procedure of the selection apparatus according to the second embodiment. As shown in FIG. 13, the selection device 200 acquires the voice data of the speakers A to E who are participants of the exchange meeting or the like as a preliminary preparation, analyzes the acquired acoustic data, and determines the acoustic characteristics of each speaker. Calculate (step S201).

Subsequently, the acquisition unit 250a of the selection device 200 acquires voice information including the voices of a plurality of speakers and stores the voice information in the voice buffer 240a (step S202).

After that, the utterance information detection unit 250b of the selection device 200 detects the utterance information 240c from the voice information (step S203).

The voice recognition unit 250c of the selection device 200 performs voice recognition on the voice information and extracts a word (step S204).

Further, the utterance impression evaluation unit 250d of the selection device 200 evaluates the participant's impression of the voice information (step S205).

Further, the praise word specifying unit 250h of the selection device 200 identifies the praise word from the words included in the utterance content character string (step S206).

Subsequently, the specifying unit 250e of the selection device 200 identifies the parrot return (step S207).

The determination unit 250f of the selection device 200 determines whether or not the impression of the speakers A to E is improved by returning the parrot (step S208).

Subsequently, the selection unit 250g of the selection device 200 evaluates the participants based on the determination result of the determination unit 250f, and stores the participant rating information 240g in the storage unit 240 (step S209). Then, the selection unit 250g selects the speaker with the highest rating as the facilitator based on the participant rating information 240g (step S210), and a series of processes is completed.

Next, the effect of the selection device 200 according to the second embodiment will be described. The selection device 200 identifies the parrot return, determines whether the participant's impression is improved by the parrot return, or whether the compliment is given together with the parrot return, and is selected as the facilitator. As a result, it is possible to select participants who can liven up the place of exchange by using praise words as well as returning the parrot as facilitators, and it is possible to liven up the exchange.

The words, utterance examples, number of utterances, situations, etc. used in the above embodiment are merely examples and can be arbitrarily changed. For example, as an example of evaluating a change in the impression of a participant, Echolalia is exemplified, but the timing is not limited to this, and a predetermined specific word (for example, a compliment, a keyword, etc.) is spoken. It can also be evaluated with.

Information including processing procedures, control procedures, specific names, various data and parameters shown in the above documents and drawings can be arbitrarily changed unless otherwise specified.

Further, each component of each of the illustrated devices is a functional concept, and does not necessarily have to be physically configured as shown in the figure. That is, the specific form of distribution or integration of each device is not limited to the one shown in the figure. That is, all or a part thereof can be functionally or physically distributed / integrated in any unit according to various loads, usage conditions, and the like.

Further, each processing function performed by each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware by wired logic.

Next, an example of a computer hardware configuration that realizes the same functions as the selection device 100 (200) shown in the above embodiment will be described. FIG. 14 is a diagram showing an example of a hardware configuration of a computer that realizes the same function as the selection device.

As shown in FIG. 14, the computer 300 has a CPU 301 for executing various arithmetic processes, an input device 302 for receiving data input from a user, and a display 303. Further, the computer 300 has a reading device 304 that reads a program or the like from a storage medium, and an interface device 305 that acquires data from a microphone, a camera, a vibration sensor, or the like via a wired or wireless network. The computer 300 has a RAM 306 that temporarily stores various information and a hard disk device 307. Then, each of the devices 301 to 307 is connected to the bus 308.

The hard disk device 307 includes an acquisition program 307a, an utterance information detection program 307b, a voice recognition program 307c, an utterance impression evaluation program 307d, a specific program 307e, a determination program 307f, and a selection program 307g. The CPU 301 reads out the acquisition program 307a, the utterance information detection program 307b, the voice recognition program 307c, the utterance impression evaluation program 307d, the specific program 307e, the determination program 307f, and the selection program 307g (further compliment specification program in the selection device 200). Expand to RAM 306.

The acquisition program 307a functions as the acquisition process 306a. The utterance information detection program 307b functions as the utterance information detection process 306b. The speech recognition program 307c functions as a speech recognition process 306c. The utterance impression evaluation program 307d functions as the utterance impression evaluation process 306d. The specific program 307e functions as the specific process 306e. The determination program 307f functions as the determination process 306f. The selection program 307g functions as the selection process 306g. In the selection device 200, the compliment identification program functions as a compliment identification process.

The processing of the acquisition process 306a corresponds to the processing of the acquisition units 150a and 250a. The processing of the utterance information detection process 306b corresponds to the processing of the utterance information detection units 150b and 250b. The processing of the voice recognition process 306c corresponds to the processing of the voice recognition units 150c and 250c. The processing of the utterance impression evaluation process 306d corresponds to the processing of the utterance impression evaluation units 150d and 250d. The processing of the specific process 306e corresponds to the processing of the specific units 150e and 250e. The processing of the determination process 306f corresponds to the processing of the determination units 150f and 250f. The processing of the selection process 306g corresponds to the processing of the selection units 150g and 250g. In the selection device 200, the processing of the compliment specifying process corresponds to the processing of the compliment specifying unit 250h.

The programs 307a to 307g do not necessarily have to be stored in the hard disk device 307 from the beginning. For example, each program is stored in a "portable physical medium" such as a flexible disk (FD), a CD-ROM, a DVD disk, a magneto-optical disk, or an IC card inserted into a computer 300. Then, the computer 300 may read and execute each program 307a to 307g.

The following additional notes will be further disclosed with respect to the embodiments including each of the above embodiments.

(Appendix 1) Voice information including the voices of a plurality of speakers is acquired, and the utterance information included in the voice information is associated with the utterance section in which the utterance is made and the utterance in the utterance section. Is detected, voice recognition is performed on the voice information, words included in the voice information are extracted, and before and after the utterance information including a specific word included in the voice information, of the plurality of speakers. A selection program characterized by evaluating an impression and causing a computer to execute a process of selecting a facilitator based on the evaluation of the impressions of the plurality of speakers.

(Appendix 2) In the utterance information adjacent to each other in chronological order, the parrot return in which the words match and the speaker is different is specified, and it is determined whether or not the parrot return improves the impression. The selection program according to Appendix 1, wherein a computer is made to execute a process of selecting the facilitator based on the result of the determination.

(Appendix 3) The selection program according to Appendix 1 or 2, wherein the evaluation process includes a process of evaluating the impression of the plurality of speakers based on the biological information of the plurality of speakers.

(Supplementary Note 4) The process of the evaluation includes the process of evaluating the impression of the plurality of speakers based on the vertical width of the pitch frequency of the voice signal included in the voice information. The selection program described in any one.

(Appendix 5) The process of identifying a compliment to another person from the word and determining the determination includes a process of determining whether or not the impression is improved by the parrot return using the compliment. The selection program described in Appendix 2, which is characterized by the above.

(Appendix 6) Voice information including the voices of a plurality of speakers is acquired, and the utterance information included in the voice information is associated with the utterance section in which the utterance is made and the utterance in the utterance section. Is detected, voice recognition is performed on the voice information, words included in the voice information are extracted, and before and after the utterance information including a specific word included in the voice information, of the plurality of speakers. A selection method characterized in that a computer executes a process of evaluating an impression and selecting a facilitator based on the evaluation of the impressions of the plurality of speakers.

(Appendix 7) Correspondence between the acquisition unit that acquires voice information including the voices of a plurality of speakers, the utterance section in which the utterance included in the voice information is performed, and the speaker who utters in the utterance section. The utterance information detection unit that detects the utterance information, the voice recognition unit that performs voice recognition on the voice information and extracts the words included in the voice information, and the voice recognition unit including the specific words included in the voice information. Selection characterized by having an utterance impression evaluation unit that evaluates the impressions of the plurality of speakers before and after the utterance information, and a selection unit that selects a facilitator based on the evaluation of the impressions of the plurality of speakers. Device.

100, 200 Detection device 110, 210 Communication unit 120, 220 Input unit 130, 230 Display unit 140, 240 Storage unit 140a, 240a Voice buffer 140b, 240b Learning acoustic feature information 140c, 240c Speech information 140d, 240d Speech impression evaluation information 140e , 240e Parrot return specific information 140f, 240f Facilitate power evaluation information 140g, 240g Participant rating information 240h Complimentary word specific information 150, 250 Control unit 150a, 250a Acquisition unit 150b, 250b Speech information detection unit 150c, 250c Voice recognition unit 150d , 250d Speech impression evaluation unit 150e, 250e Specific unit 150f, 250f Judgment unit 150g, 250g Selection unit 250h Complimentary word identification unit

Claims (7)

Acquires voice information that includes the voices of multiple speakers,
The utterance information in which the utterance section in which the utterance was made and the speaker who made the utterance in the utterance section are associated with each other is detected.
Voice recognition is performed on the voice information, words included in the voice information are extracted, and the words are extracted.
Impressions of the plurality of speakers are evaluated before and after the utterance information including a specific word included in the voice information.
A selection program characterized by having a computer execute a process of selecting a facilitator based on the evaluation of the impressions of the plurality of speakers. In the utterance information adjacent to each other in chronological order, the word is matched and the speaker identifies a different parrot return.
It is determined whether or not the impression is improved by the parrot return, and it is determined.
The selection program according to claim 1, wherein a computer executes a process of selecting the facilitator based on the result of the determination. The selection program according to claim 1 or 2, wherein the evaluation process includes a process of evaluating the impression of the plurality of speakers based on the biological information of the plurality of speakers. One of claims 1 to 3, wherein the evaluation process includes a process of evaluating the impression of the plurality of speakers based on the vertical width of the pitch frequency of the voice signal included in the voice information. The selection program described in one. Identify words of praise to others from the above words,
The selection program according to claim 2, wherein the determination process includes a process of determining whether or not the impression is improved by the parrot return using the compliment. Acquires voice information that includes the voices of multiple speakers,
The utterance information in which the utterance section in which the utterance was made and the speaker who made the utterance in the utterance section are associated with each other is detected.
Voice recognition is performed on the voice information, words included in the voice information are extracted, and the words are extracted.
Impressions of the plurality of speakers are evaluated before and after the utterance information including a specific word included in the voice information.
A selection method characterized in that a computer executes a process of selecting a facilitator based on the evaluation of the impressions of the plurality of speakers. An acquisition unit that acquires voice information that includes the voices of multiple speakers, and
An utterance information detection unit that detects utterance information in which the utterance section in which the utterance is performed included in the voice information and the utterance speaker in the utterance section are associated with each other.
A voice recognition unit that performs voice recognition on the voice information and extracts words included in the voice information.
An utterance impression evaluation unit that evaluates the impressions of the plurality of speakers before and after the utterance information including a specific word included in the voice information.
A selection unit that selects facilitators based on the evaluation of the impressions of multiple speakers,
A selection device characterized by having.

Images