JP2023055296A - Compatibility assessment system, method, and program - Google Patents

Abstract

To provide a compatibility assessment system capable of supporting a reduced turnover rate in a workplace.SOLUTION: A compatibility assessment system comprises: an acquisition unit 32 for acquiring utterances by an utterer; a detection unit 33 for detecting a degree of psychological safety for each utterance; and an assessment unit 34 for assessing compatibility for each pair between a target utterer and another utterer who utters immediately before the utterance of the target utterer, based on the degree detected for the utterance of the target utterer.SELECTED DRAWING: Figure 4

Description

The present disclosure relates to a system, method, and program for compatibility evaluation.

WO2019/097674 discloses a human relationship analysis unit that analyzes and quantifies human relationships. The human analysis unit extracts a predetermined keyword from conversation voice data of a plurality of people, identifies the wording of each speaker based on the keyword of each speaker, and uses the wording and conversation content to analyze the human Analyze relationships.

WO2019/097674

In general, it is desirable to have a low turnover rate in the workplace. The most common reason for leaving a job is said to be a problem with human relationships. In other words, the turnover rate depends on the compatibility between a plurality of people in the workplace.

The human relationship analysis unit disclosed in Patent Document 1 is applied to a vehicle operation support device such as a car navigation system, and the human relationships among a plurality of occupants in the vehicle are "family," "friends," and "supervisor and subordinate." , "colleagues", etc. That is, the human relationship analysis unit disclosed in Patent Document 1 does not estimate the compatibility between a plurality of people. Therefore, the human relationship analysis unit disclosed in Patent Document 1 cannot help reduce the turnover rate in the workplace.

The present disclosure has been made to solve the above-mentioned problems, and its purpose is to provide a compatibility evaluation system, method, and program that can help reduce the turnover rate in the workplace. be.

According to one aspect, the compatibility evaluation system includes an acquisition unit that acquires a statement by a speaker, a detection unit that detects the degree of psychological safety for each statement, and , and an evaluation unit that evaluates compatibility for each pair between the target speaker and another speaker who spoke immediately before the target speaker's speech.

Preferably, the evaluation unit calculates the evaluation value of compatibility for each pair by totaling the degree of detection of the utterances of the target speaker for each pair for each predetermined period.

Preferably, the compatibility evaluation system further includes a providing unit that provides a screen showing changes in evaluation values over time.

Preferably, the evaluation value includes a first numerical value indicating the number of utterances whose degree exceeds the first criterion, a second numerical value indicating the percentage of utterances whose degree exceeds the first criterion, and the number of utterances whose degree falls below the second criterion. of the third numerical value shown, the fourth numerical value indicating the percentage of utterances whose degree falls below the second standard, and the fifth numerical value calculated using one or more numerical values selected from the first numerical value to the fourth numerical value At least one.

Preferably, the compatibility evaluation system includes a warning unit that issues a warning indicating a sign of deterioration of compatibility in response to a change in evaluation value in the direction of worsening compatibility for each pair in the past fixed period exceeding a reference value. further provide.

Preferably, the compatibility evaluation system further includes a labeling unit that labels the target speaker based on the temporal change in the evaluation value for each pair.

Preferably, the labeling unit calculates, for each pair, a first change range of the evaluation value in the direction of improving compatibility in a first fixed period after the start of compatibility evaluation. The labeling part indicates that the first representative value of the first change width for each pair exceeds the second reference value and the number of pairs for which compatibility evaluation is started per unit time exceeds the first reference value. to assign the first label to the target speaker.

Preferably, the labeling unit calculates, for each pair, a second variation range of the evaluation value in a direction of improving compatibility in a second fixed period after the start of compatibility evaluation, A second label is attached to the target speaker in accordance with the fact that the second representative value of the width of change exceeds the third reference value.

Preferably, the labeling unit calculates a third representative value of the evaluation values for each pair in the past fixed period, and the third representative value is included in a target range that is less compatible than the median value in the range that the evaluation values can take. A third label is attached to the target speaker according to the

Preferably, the speaker is a worker. The compatibility evaluation system further includes a database that manages the working hours of each speaker. The labeling unit calculates, for each pair, the third variation range of the evaluation value in the past fixed period in the direction of worsening compatibility. The labeling unit, in response to the fourth representative value of the third change width for each pair exceeding the fourth reference value and the working hours of the target speaker exceeding the fifth reference value, to the target speaker: Attach a fourth label.

Preferably, the detection unit acquires the degree by inputting data indicating the utterance to the trained model. The trained model is generated by machine learning using teacher data so that when data indicating a statement is input, the degree of psychological safety of the person who made the statement is output.

Preferably, the acquisition unit acquires at least one of voice data, e-mail data, and message data posted in chat as the data indicating the utterance.

According to another aspect, a method includes the steps of obtaining utterances by a speaker; detecting a degree of psychological safety for each utterance; Evaluating the compatibility of each pair between the speaker and another speaker who spoke immediately before the target speaker's speech.

According to yet another aspect, a program causes a computer to perform the above method.

The present disclosure can help reduce turnover in the workplace.

1 is a diagram showing a schematic configuration of a conference system according to an embodiment; FIG. 3 is a diagram showing a hardware configuration of a terminal according to this embodiment; FIG. It is a figure which shows the hardware constitutions of the information processing apparatus which concerns on this Embodiment. It is a figure which shows typically the software configuration of an information processing apparatus. It is a figure which shows an example of a basic database. It is a figure which shows an example of a work database. It is a figure which shows an example of a summary database. It is a figure which shows the example of a process of a labeling part. FIG. 10 is a diagram showing an example of a screen showing temporal changes in evaluation values; FIG. 10 is a diagram showing another example of a screen showing temporal changes in evaluation values; FIG. 10 is a diagram showing an example of a screen showing a list of labels; 10 is a flow chart showing the flow of update processing of a summary database; FIG. 10 is a flow chart showing a flow of determination processing of a pair to be warned; FIG. 4 is a flowchart showing the flow of labeling processing;

Hereinafter, embodiments and modifications according to the present invention will be described with reference to the drawings. In the following description, identical parts and components are given identical reference numerals. Their names and functions are also the same. Therefore, detailed description of these will not be repeated. Note that the embodiments and modifications described below may be selectively combined as appropriate.

<A. Psychological Safety>
This embodiment provides an evaluation tool for evaluating the compatibility of each pair of two members by detecting the degree of psychological safety of the members participating in the meeting.

A method that focuses on "psychological safety" has been proposed as a method of reducing the psychological barriers during meetings for the purpose of activating discussions and improving communication in companies. Psychological safety refers to the state of reducing the psychological barriers of members in a meeting (for example, the state of being confident that other members will not reject their remarks or punish them). Professor Edmonson's "Seven Questions" is well known as an item for measuring psychological safety.

Professor Edmonson's "Seven Questions" are:
Question (1) When you make a mistake in a team, you are usually blamed.
Question (2) Team members can point out challenges and difficult problems.
Question (3) Team members sometimes reject others because they are different from themselves.
Question (4): Is it safe to take risky actions towards the team?
Question (5) It is difficult to ask other members of the team for help.
Question (6) No member of the team intentionally degrades the work of others.
Question (7) When working with team members, do I feel that my skills and talents are respected and put to good use?

The higher the probability of the conditions shown in questions (1), (3), and (5), the lower the degree of psychological safety. On the other hand, the higher the probability of the states shown in questions (2), (4), (6), and (7), the higher the degree of psychological safety.

In addition, Professor Edmonson suggests that low psychological safety triggers four types of anxiety. The four fears are "fear of perceived ignorance,""fear of perceived incompetence,""fear of being disturbed," and "fear of perceived negative." For example, it is assumed that these four concerns are resolved by the following remarks.
・Constructive criticism and different opinions were expressed in response to other people's remarks (it is assumed that negative concerns have been resolved).
・They confided in their own problems, anxieties, and mistakes (it is assumed that the incompetence of incompetence has been resolved).
- Asked for help from others (assumed that fears of being incompetent and disturbing were resolved).
・New proposals and opinions were expressed (it is assumed that the anxiety that seems to be disturbing has been resolved).
・I asked questions about things I didn't know (it is assumed that the fear of being ignorant has been resolved).

Conversely, remarks that express only critical opinions tend to be made in an environment where the four anxieties have not been resolved.

Therefore, in the present embodiment, the degree of psychological safety is detected based on remarks related to the above questions (1) to (7) and remarks related to four anxiety by members participating in the meeting. be.

<B. System>
FIG. 1 is a diagram showing a schematic configuration of a conference system according to this embodiment. A system 1 shown in FIG. 1 is a system for providing an online conference in which images and voices are exchanged via a network. As shown in FIG. 1, a system 1 includes a plurality of terminals 200 that can be operated by members (participants) participating in a conference, and an information processing device 300 that communicates with the plurality of terminals 200 via a network 100. . Below, the online conference is abbreviated as “meeting”.

Terminal 200 includes, for example, a general-purpose PC (Personal Computer), a smartphone, or a tablet. The information processing device 300 provides a platform for implementing a conference, for example, as a cloud server with the terminal 200 as a client. The network 100 may be, for example, a public network such as the Internet, a public line, or a public wireless LAN (Local Area Network), or may be a private network such as a LAN or VPN (Virtual Private Network). The information processing device 300 is illustrated as a cloud server, but is not limited to this, and may be an on-premises server.

The information processing device 300 supports online conferences among a plurality of terminals 200 . The information processing apparatus 300 receives, for example, video data, audio data, etc. sent from one terminal 200 and transfers the received video data or audio data to another terminal 200 . Thereby, the members can operate the terminal 200 to share the video and audio.

Furthermore, the information processing device 300 detects the degree of psychological safety of the members during the meeting from the voice data exchanged between the terminals 200 in the meeting, and evaluates the compatibility of each pair consisting of two members. provide assessment tools for

<C. Hardware configuration>
(Device hardware configuration)
FIG. 2 is a diagram showing the hardware configuration of the terminal according to this embodiment. As shown in FIG. 2, the terminal 200 includes a processor 11, a main memory 12 composed of a volatile storage medium including, for example, a RAM (Random Access Memory), a camera 13, a microphone 14, A communication interface 15 composed of a communication circuit including a NIC (Network Interface Card), an input interface 16 connecting an input device 16A such as a user-operable keyboard and mouse, a display interface 17 connecting a display 17A, A memory composed of a speaker 18, an optical drive 19 to which an optical disc 19A, which is an example of a recording medium, is detachably mounted, and a non-volatile recording medium such as a HDD (Hard Disk Drive) or an SSD (Solid State Drive). device 20; These components are communicatively connected to each other via an internal bus 25 . The input interface 16 receives user operations on the terminal 200 via the input device 16A. Display interface 17 also includes a driver circuit that drives display 17A based on the image data to display an image. The storage device 20 stores a program 22 that implements various processes when read and executed by the processor 11 .

Processor 11 is configured by, for example, at least one integrated circuit. The integrated circuit is composed of, for example, at least one CPU (Central Processing Unit), at least one ASIC (Application Specific Integrated Circuit), at least one FPGA (Field Programmable Gate Array), or a combination thereof.

The processor 11 develops the program 22 stored in the storage device 20 in the main memory 12 and executes it, thereby realizing various processes according to the present embodiment. The main memory 12 provides work memory necessary for program execution by the processor 11 .

The camera 13 is an imaging device having a CMOS (Complementary Metal Oxide Semiconductor) sensor, a CCD (Charge Coupled Device) sensor, or the like. The imaging field of view of the camera 13 has a range capable of imaging members operating the terminals 200, and generates image data such as captured moving images (video) or still images. Microphone 14 is configured to collect sounds around terminal 200 . The sound collected by the microphone 14 may include the sound of the member operating the terminal 200 .

The communication interface 15 exchanges data and signals with the information processing device 300 via the network 100 . As an example, the terminal 200 transmits and receives image data, audio data, etc. to and from the information processing device 300 via the communication interface 15 . The image data includes image data based on an image captured by the camera 13 of the terminal 200 , and the audio data includes audio data collected by the microphone 14 of the terminal 200 .

During the conference, the information processing device 300 receives image data and audio data transmitted from the terminal 200 and transfers the received image data and audio data to the terminals 200 of other members. The terminal 200 displays an image based on the image data received from the information processing device 300 on the display 17A, and outputs sound based on the audio data received from the information processing device 300 from the speaker 18 . As a result, during the conference, the members can communicate with each other by exchanging images and voices through the terminals 200 .

The communication interface 15 can also receive screen data from the information processing device 300 . More specifically, information processing apparatus 300 serves as a web server and transmits a web page to terminal 200, and the web browser of terminal 200 displays a screen based on the web page from information processing apparatus 300 via display interface 17. It is displayed on the display 17A.

The optical drive 19 reads various programs (for example, the program 22 ) stored therein from an external storage medium such as an optical disc 19 A and installs them in the storage device 20 . FIG. 2 shows a configuration example in which a necessary program is installed on the terminal 200 from an external storage medium. Also, the external storage medium is not limited to the optical disc 19A, and may be a storage medium such as a USB (Universal Serial Bus) memory, an SD (Secure Digital) card, or a CF (Compact Flash).

The program 22 in the storage device 20 may be provided not as a single program but as part of any program. In this case, processing according to the present embodiment is realized in cooperation with an arbitrary program. Even a program that does not include some of such modules does not deviate from the gist of terminal 200 according to the present embodiment. Also, part or all of the functions provided by the program 22 may be realized by dedicated hardware. Also, the camera 13 , the microphone 14 and the speaker 18 may be externally attached to the terminal 200 , and these devices may be connected to the terminal 200 without being limited to one device.

(Hardware configuration of information processing device)
FIG. 3 is a diagram showing the hardware configuration of the information processing apparatus according to this embodiment. As shown in FIG. 3, an information processing apparatus 300 includes a processor 301, a main memory 302 composed of a volatile storage medium including a RAM, etc., a communication interface 303 composed of a communication circuit including an NIC, a user An input interface 305 for connecting an input device 305A such as a keyboard or a mouse that can be operated by , a display interface 306 for connecting a display 306A, and an optical drive 307 for detachably mounting an optical disk 307A, which is an example of a recording medium. , and a storage device 310 composed of a non-volatile recording medium such as an HDD or SSD. These components are communicatively connected to each other via an internal bus 319 . Input interface 305 receives user operations on information processing apparatus 300 via input device 305A. Display interface 306 also includes driver circuitry that drives display 306A based on the image data to display an image. Storage device 310 stores software that is read and executed by processor 11 . This software consists of a program 320 that implements various processes, a DB group 330 consisting of a plurality of databases (data bases), and a trained model 340 that receives input of voice data and outputs the degree of psychological safety. ,including.

Processor 301 is composed of, for example, at least one integrated circuit. An integrated circuit is composed of, for example, at least one CPU, at least one ASIC, at least one FPGA, or a combination thereof.

Processor 301 expands program 320 stored in storage device 310 into main memory 302 and executes it, thereby realizing various processes according to the present embodiment. The main memory 302 provides work memory required for program execution by the processor 301 .

Communication interface 303 exchanges data and signals with terminal 200 via network 100 . As an example, the information processing device 300 transmits and receives image data, audio data, screen data, etc. to and from the terminal 200 via the communication interface 303 .

The optical drive 307 reads various programs (for example, the program 320 ) and data such as DBs stored therein from an optical disc 307 A or the like that is an external storage medium, and installs them in the storage device 310 . FIG. 3 shows a configuration example in which necessary programs are installed in the information processing apparatus 300 via the optical drive 307. However, the present invention is not limited to this, and may be downloaded from another apparatus on the network 100. good. Also, the external storage medium is not limited to the optical disk 307A, and may be a storage medium such as a USB memory, an SD card, or a CF.

It should be noted that the program 320 may be provided as a part of an arbitrary program, not as a standalone program. In this case, processing according to the present embodiment is realized in cooperation with an arbitrary program. Even a program that does not include such a part of modules does not deviate from the gist of information processing apparatus 300 according to the present embodiment. Also, part or all of the functions provided by the program 320 may be realized by dedicated hardware.

<D. Software Configuration of Information Processing Device>
FIG. 4 is a diagram schematically showing the software configuration of the information processing device. FIG. 4 shows the software configuration associated with the assessment tool for assessing pairwise compatibility of two members. That is, FIG. 4 omits the software configuration for supporting the conference. Note that the information processing apparatus 300 may have a known software configuration for supporting the conference.

As shown in FIG. 4, the information processing device 300 includes a storage unit 30, a model generation unit 31, an acquisition unit 32, a detection unit 33, an evaluation unit 34, a provision unit 35, a warning unit 36, and a labeling portion 37 . Storage unit 30 is implemented by main memory 302 and storage device 310 shown in FIG. The model generation unit 31, the acquisition unit 32, the detection unit 33, the evaluation unit 34, the provision unit 35, the warning unit 36, and the labeling unit 37 are implemented by executing the program 320 by the processor 301 shown in FIG.

(storage unit)
The storage unit 30 stores a learned model 340 , one or more basic databases 331 , a summary database 332 , and a work database 333 . The trained model 340 is generated by the model generation unit 31 and stored in the storage unit 30 . The basic database 331 is generated by the acquisition unit 32 and the detection unit 33 and stored in the storage unit 30 . The summary database 332 is updated by the evaluation section 34 . The work database 333 is updated by the acquisition unit 32 .

(model generator)
The model generation unit 31 generates a learned model 340 by performing machine learning using the teacher data group, and stores the generated learned model 340 in the storage unit 30 .

The training data group includes, for example, speech data collected in past meetings and the degree of psychological safety given to the speech indicated by the speech data (hereinafter referred to as "psychological safety score"). ) are associated with each other.

The psychological safety score may be set to increase as the psychological safety increases, or may be set to decrease as the psychological safety increases.

The trained model 340 is defined by computational parameters used in the computational process of extracting features from audio data and generating psychological safety scores from the extracted features. In machine learning, when a training data group is given, calculation parameters are set so that when voice data included in the training data group is input, a psychological safety score associated with the voice data is output. value is adjusted. The features extracted from the voice data include, for example, characters, morphemes, words, parameters indicating intonation (for example, frequency), and the like. Characters, morphemes and words are extracted from the speech data using known speech recognition techniques and natural language processing.

The trained model 340 is, for example, a neural network such as a recurrent neural network (RNN: Recurrent Neural Network), a support vector machine (SVM: Support Vector Machine), or a machine learning model such as CRF (Conditional Random Fields). . When generating the trained model 340 configured by a neural network, the calculation parameters adjusted by machine learning include weights of connections between neurons included in each layer, threshold values set for each neuron, and the like. In addition, the number of layers, the number of neurons in each layer, the connection between neurons in adjacent layers, the initial weight of the connection between each neuron, the initial threshold value of each neuron, etc. are given by the template. , or may be given by user input.

(acquisition unit)
For each conference, the acquisition unit 32 acquires the speech from the terminal 200 of the speaker in the conference. Specifically, the acquisition unit 32 creates the basic database 331 for each conference, and adds data to the basic database 331 based on the voice data acquired from the terminal 200 of the speaker of the conference. The acquisition unit 32 stores the created basic database 331 in the storage unit 30 .

FIG. 5 is a diagram showing an example of the basic database. The basic database 331 associates a personal ID that identifies a speaker, a speech time, voice data, and a psychological safety score. As shown in FIG. 5, the basic database 331 has a tabular format and includes one or more records 331a. Each record 331a includes a field 331b describing an individual ID, a field 331c describing the time of speech, a field 331d describing a path to voice data, and a field 331e describing a psychological safety score. including.

The acquiring unit 32 adds a new record 331a after acquiring the voice data. The acquiring unit 32 writes information related to the acquired audio data in the field of the new record 331a. Specifically, the acquisition unit 32 writes in the field 331b the personal ID of the speaker who uses the terminal 200 that is the output source of the voice data. The acquisition unit 32 describes the acquisition time of the voice data in the field 331c as the utterance time. After storing the acquired audio data in the storage unit 30, the acquisition unit 32 describes the path to the audio data in the field 331d. The acquisition unit 32 writes nothing in the field 331e.

Furthermore, the acquisition unit 32 acquires attendance information from an attendance management server (not shown), and updates the work database 333 based on the acquired attendance information.

FIG. 6 is a diagram showing an example of a work database. The working database 333 manages the working hours of each worker for each period. As shown in FIG. 6, the work database 333 associates a worker's personal ID, a period, and attendance information for that period. Attendance information includes attendance days, overtime hours, working hours, absence days, and the like.

(Detection unit)
The detection unit 33 detects the degree of psychological safety (psychological safety score) for each statement. The detection unit 33 may use the trained model 340 to detect the psychological safety score for each statement. Specifically, for each record 331a (see FIG. 5) of the basic database 331, the detection unit 33 inputs the speech data stored in the path described in the field 331d to the learned model 340, and the learned model 340 Get the psychological safety score output from . The detection unit 33 writes the obtained psychological safety score in the field 331e.

(Evaluation Department)
The evaluation unit 34 evaluates the compatibility of each pair between the evaluation subject and another speaker who spoke immediately before the evaluation subject's utterance, based on the psychological safety score detected for the evaluation subject's utterance. . Specifically, the evaluation unit 34 calculates the compatibility of each pair by totaling the psychological safety scores detected in the statements of the persons to be evaluated in each pair for each predetermined period (for example, one month). Calculate the evaluation value for The evaluation unit 34 updates the tally database 332 according to the tally result.

FIG. 7 is a diagram showing an example of a summary database. As shown in FIG. 7, the summary database 332 has a tabular format and includes one or more records 332a. Each record 331a includes a field 332b in which the personal ID of the person to be evaluated (speaker) is described, a field 332c in which the personal ID of the immediately preceding speaker is described, a field 332d in which the period is described, and various evaluations. It includes fields 332e-332h in which values are described and a field 332i in which total talk time is described. The field 332e describes the number of statements with high psychological safety. Field 332f describes the ratio of utterances with high psychological safety. The field 332g describes the number of statements with low psychological safety. Field 332h describes the rate of utterances with low psychological safety.

After a predetermined period (for example, one month) elapses, the evaluation unit 34 selects the basic database 331 describing the utterance times belonging to the period from among the one or more basic databases 331 stored in the storage unit 30. Extract. The evaluation unit 34 rearranges the records 331a of each of the extracted basic databases 331 (see FIG. 5) in ascending order of the utterance time described in the field 331c.

The evaluation unit 34 selects one person to be evaluated from among the participating members of the conference. The evaluation unit 34 identifies, from the extracted basic database 331, a record 331a in which the personal ID of the person to be evaluated is described in the field 331b (see FIG. 5). The evaluation unit 34 selects the personal ID described in the field 331b of the record 331a immediately before the specified record 331a as that of the speaker who spoke immediately before the evaluation subject (hereinafter referred to as the "previous speaker"). Identify as personal ID. The evaluation unit 34 adds a new record 332a (see FIG. 7) corresponding to the pair of the person to be evaluated and the previous speaker to the aggregation database 332 for each previous speaker.

In the added record 332a, the evaluation unit 34 describes the personal ID of the person to be evaluated, the personal ID of the previous speaker, and the period in fields 332b, 332c, and 332d, respectively.

The evaluation unit 34 classifies one or more records 331a specified from the basic database 331 according to the previous speaker. The evaluation unit 34 calculates various evaluation values by comparing the psychological safety score described in the field 331e of the record 331a classified for each previous speaker with a predetermined threshold. For example, if the setting is such that the higher the psychological safety, the higher the psychological safety score, the evaluation unit 34 counts the number of records 331a whose psychological safety score exceeds the first threshold as the psychological safety score. counted as the number of high-quality remarks. Furthermore, the evaluation unit 34 counts the number of records 331a whose psychological safety score is below the second threshold (<first threshold) as the number of statements with low psychological safety. Alternatively, if the setting is such that the higher the psychological safety, the smaller the psychological safety score, the evaluation unit 34 counts the number of records 331a whose psychological safety score is below the third threshold as the psychological safety score. counted as the number of high-quality remarks. Furthermore, the evaluation unit 34 counts the number of records 331a whose psychological safety score exceeds the fourth threshold (>third threshold) as the number of statements with low psychological safety.

The evaluation unit 34 calculates the percentage of statements with high psychological safety by dividing the number of statements with high psychological safety by the number of records 331a classified for each previous speaker. The evaluation unit 34 calculates the percentage of statements with high psychological safety by dividing the number of statements with low psychological safety by the number of records 331a classified for each previous speaker.

The evaluation unit 34 evaluates the number of utterances with high psychological safety, the ratio of utterances with high psychological safety, the number of utterances with low psychological safety, and the number of utterances with low psychological safety obtained in this way. Speech percentages are described in fields 332e-332h, respectively.

Various evaluation values calculated by the evaluation unit 34 relate to compatibility of the pair. For example, the greater the number of statements with high psychological safety, the better the compatibility of the pair. The higher the percentage of utterances with high psychological safety, the better the compatibility of the pair. Conversely, the greater the number of utterances with low psychological safety, the worse the compatibility of the pair. The higher the percentage of utterances with low psychological safety, the worse the compatibility of the pair. Therefore, by checking the evaluation value, it is possible to grasp the compatibility of each pair. Workers in pairs with long-term incompatibility are more likely to quit. Therefore, according to the compatibility of each pair, for example, interviews or personnel transfers can be conducted, thereby suppressing a decrease in the turnover rate.

The evaluation unit 34 calculates the total speech time indicated by the speech data corresponding to the record 331a classified by the immediately preceding speaker as the total conversation time. The evaluation unit 34 writes the calculated total conversation time in the field 332i. In general, the longer the total conversation time, the better the compatibility of the pair.

(providing department)
The providing unit 35 provides a screen showing evaluation results by the evaluating unit 34 . For example, the providing unit 35 provides a screen showing temporal changes in evaluation values. The screen provided by the providing unit 35 is displayed on the display 306A.

(warning part)
The warning unit 36 indicates a sign of worsening compatibility in response to a change in the evaluation value of the target pair in the direction of worsening compatibility over a certain past period (for example, half a year, one year, etc.) exceeding a reference value. issue a warning that indicates The warning is displayed on the screen provided by the providing unit 35 .

(Labeling part)
The labeling unit 37 selects one person to be labeled from among the participating members of the conference. The labeling unit 37 labels the person to be labeled based on the temporal change in the evaluation value for each pair of the person to be labeled (speaker) and the previous speaker. For example, if the aggregation database 332 has two records 332a, a pair of the labeling target person and "Mr. A" and a pair of the labeling target person and "Mr. B", the labeling unit 37 A label is determined based on the time change of the evaluation value. A label expresses the character and psychological state of a person to be labeled. The label given by the labeling unit 37 is displayed on the screen provided by the providing unit 35 .

FIG. 8 is a diagram illustrating a processing example of the labeling unit; In FIG. 8, the left column shows four examples of graphs showing temporal changes in evaluation values. FIG. 8 shows an evaluation value that increases as psychological safety increases (for example, the percentage of statements with high psychological safety). The right column shows labels corresponding to the graphs shown in the left column.

((Method of attaching the first label))
As shown in the first row of FIG. 8, the labeling unit 37 calculates the evaluation value in the direction of improving compatibility for each pair in a certain period of time T1 (for example, three months) after the start of compatibility evaluation. A change width W1 is calculated. The timing at which the evaluation unit 34 starts evaluating compatibility is the timing at which the recording of the evaluation values in the aggregation database 332 is started, and is usually close to the timing at which the person to be labeled and the previous speaker get to know each other. Therefore, the fixed period T1 is a period starting from the timing when the person to be labeled and the previous speaker get to know each other.

The labeling unit 37 compares the representative value (for example, average value, median value, minimum value, etc.) of the change width W1 for each pair with the first reference value. The first reference value is determined in advance from the value of the width of change W1 for people with whom human relations are well established. When the representative value of the change width W1 for each pair exceeds the first reference value, it means that the compatibility with the newly acquainted person will increase in a short period of time.

Further, the labeling unit 37 compares the number of pairs whose compatibility evaluation is started per unit time (for example, one month) with the second reference value. The second reference value is determined in advance based on the number of people with whom good human relations are acquainted per unit time. When the number of pairs for which compatibility evaluation is started per unit time exceeds the second reference value, it means that there are many new acquaintances.

The labeling unit 37, when the representative value of the change width W1 for each pair exceeds the first reference value and the number of pairs for which compatibility evaluation is started per unit time exceeds the second reference value, A label "Good relationship building" (first label) is attached to the labeling target.

((Secondary labeling method))
As shown in the second row of FIG. 8, the labeling unit 37 calculates the evaluation value in the direction of improving compatibility for each pair in a certain period of time T2 (for example, 6 months) after the start of compatibility evaluation. A change width W2 is calculated. The fixed period T2 is longer than the fixed period T1.

The labeling unit 37 compares the representative value (for example, average value, median value, minimum value, etc.) of the change width W2 for each pair with the third reference value. The third reference value is predetermined from the value of the positive person's change width W2. When the representative value of the change width W2 for each pair exceeds the third reference value, it means that the compatibility with a newly acquainted person is high over a long period of time, although the person is not good at building human relationships.

The labeling unit 37 attaches the label “positive” (second label) to the person to be labeled when the representative value of the change width W2 for each pair exceeds the third reference value.

((Method of attaching the third label))
As shown in the third row of FIG. 8, the labeling unit 37 monitors the evaluation value for each pair during a certain past period T3 (for example, one year going back from the present). The labeling unit 37 calculates a representative value (e.g., average value, median value, maximum value) of the evaluation values for each pair in the past fixed period T3, and selects an object that is less compatible than the median value within the range of possible evaluation values. It is determined whether or not the representative value is included in the range R. The target range R is predetermined so as to include evaluation values that a negative person can take. The labeling unit 37 attaches the label “negative” (third label) to the person to be labeled according to the fact that the representative value of the evaluation values is included in the target range R.

((Method of attaching the fourth label))
As shown in the fourth row of FIG. 8, the labeling unit 37 indicates, for each pair, the change width W3 of the evaluation value in the direction of worsening compatibility during a certain past period T4 (for example, one year before the present). calculate.

The labeling unit 37 compares the representative value (for example, average value, median value, minimum value, etc.) of the change width W3 for each pair with the fourth reference value. The fourth reference value is predetermined, for example, from the value of the change width W3 of a person who has left a job in the past or a disease (especially a mental disease). If the representative value of the change width W3 for each pair exceeds the fourth reference value, it means that the psychological safety is low and the possibility of leaving the job or getting sick is high.

Further, the labeling unit 37 reads the working hours of the person to be labeled from the work database 333, and compares the read working hours with the fifth reference value. The fifth reference value is determined in advance based on the working hours of the person who left the job in the past or the person who got sick. Working hours exceeding the fifth reference value mean that psychological safety is low and the possibility of quitting the job or getting sick is high.

The labeling unit 37 assigns the label “Separation/Sickness” to the person to be labeled in response to the fact that the representative value of the change width W3 for each pair exceeds the fourth reference value and the working hours exceed the fifth reference value. "Signs of" (4th label).

<E. Screen example>
Next, examples of screens provided by the providing unit 35 will be described. For example, the providing unit 35 provides a screen showing temporal changes in evaluation values. Alternatively, the providing unit 35 may provide a screen showing a list of labels given to members belonging to a certain department. A screen provided by the providing unit 35 is checked by, for example, a personnel manager or an administrator.

(Screen showing changes over time in evaluation values)
FIG. 9 is a diagram showing an example of a screen showing temporal changes in evaluation values. FIG. 10 is a diagram showing another example of a screen showing temporal changes in evaluation values. As shown in FIGS. 9 and 10 , screen 50 includes area 51 , entry fields 52 and 53 , warning indicator 54 and label display field 55 .

The input field 52 accepts the input of the person to be evaluated. The input field 53 accepts the input of the person to be evaluated. For example, the person in charge of human resources or the manager can select the person to be evaluated and the person to be evaluated from the pull-down list.

The providing unit 35 retrieves from the tally database 332 (see FIG. 7) a field 332b in which the personal ID of the person to be evaluated entered in the input field 52 and a field 332b in which the personal ID of the other party entered in the input field 53 is entered. Extract the record 332a described in 332c. Based on the period described in field 332d of the extracted record 332a and the evaluation values described in fields 332e to 332h, the providing unit 35 creates a graph showing temporal changes in evaluation values.

A region 51 displays a graph showing changes in evaluation values over time. In the examples shown in FIGS. 9 and 10 , a graph showing temporal changes in the ratio of statements with high psychological safety is displayed in area 51 .

The warning unit 36 detects a change in the evaluation value of the person to be evaluated and the target pair of the partner input in the input fields 52 and 53 in the direction of worsening compatibility over a certain period of time in the past (for example, half a year, one year, etc.). Compare the width with a reference value. If the width of change exceeds the reference value, the warning unit 36 instructs the providing unit 35 to turn on the warning indicator 54 . In the example shown in FIG. 10, the evaluation value (proportion of statements with high psychological safety) displayed in area 51 tends to decrease, and warning indicator 54 is lit. On the other hand, when the width of change is equal to or less than the reference value, the warning unit 36 instructs the providing unit 35 to turn off the warning indicator 54 . In the example shown in FIG. 9, the evaluation value (proportion of statements with high psychological safety) displayed in area 51 is on the rise, and warning indicator 54 is turned off.

By checking the warning indicator 54, the person in charge of human resources or the manager can easily grasp that the compatibility of the target pair is deteriorating. In response to the warning indicator 54 lighting, the person in charge of human resources or the manager interviews the person to be evaluated or the other party, or makes sure that the person to be evaluated and the other party do not belong to the same team. can. Alternatively, the HR person or manager may consider moving one of the appraiser and the other.

In the label display column 55, the label given to the evaluation subject input in the input column 52 is displayed. The person in charge of human resources or the manager can grasp the character and psychological state of the person to be evaluated by checking the label display field 55 .

(Screen showing a list of labels)
FIG. 11 is a diagram showing an example of a screen showing a list of labels. As shown in FIG. 11, screen 60 includes an input field 61 and display fields 62 and 63 .

The input field 61 accepts the input of the department to be evaluated. For example, a person in charge of human resources or a manager can select a department to be evaluated from a pull-down list.

The warning unit 36 determines the degree of change in the evaluation value in the past fixed period (for example, half a year, one year, etc.) in the direction of worsening compatibility and Compare with reference value. The warning unit 36 displays in the display field 62 a list of pairs whose change width exceeds the reference value. In other words, the display field 62 displays the pair that is the target of the warning indicating the sign of deterioration of compatibility.

By checking the display column 62, the person in charge of personnel affairs or the manager can easily grasp the pair showing signs of deterioration of compatibility. The person in charge of personnel affairs or the manager can interview the member of the pair displayed in the display field 62, or take care that the pair does not belong to the same team.

The display field 63 displays the label assigned to each member of the evaluation target department input in the input field 61 . By checking the display column 63, the person in charge of human resources or the manager can grasp the character and psychological state of each member.

<F. Processing Flow>
The flow of processing by the processor 301 of the information processing apparatus 300 will be described with reference to FIGS. 12 to 14. FIG.

(database update)
FIG. 12 is a flow chart showing the flow of update processing of the summary database. First, processor 301 determines whether a conference has started (step S1). If the conference has not started (NO in step S1), the processor 301 advances the process to step S7.

If the conference has started (NO in step S1), processor 301 creates basic database 331 corresponding to the conference (step S2). Next, the processor 301 acquires statements made by speakers in the conference (step S3). Specifically, the processor 301 acquires voice data from the terminals 200 used by the conference participants.

Next, processor 301 detects the degree of psychological safety (psychological safety score) for the acquired statement (step S4). Specifically, the processor 301 obtains a psychological safety score output from the trained model 340 by inputting audio data to the trained model 340 .

Next, the processor 301 adds a new record 331a to the basic database 331 and writes information in each field of the record 331a (step S5). Specifically, the processor 301 writes in the field 331b the personal ID of the speaker using the terminal 200 that transmitted the voice data acquired in step S3, and writes the time at which the voice data was acquired in the field 331c. In addition, processor 301 describes the path to the storage location of the audio data in field 331d and the psychological safety score in field 331e.

Next, processor 301 determines whether the conference has ended (step S6). If the conference has not ended (NO in step S6), processor 301 returns the process to step S3. If the conference has ended (YES in step S6), processor 301 advances the process to step S7.

In step S7, the processor 301 determines whether or not the current time is the time to start counting. For example, if the aggregation period is set to one month, the processor 301 may determine that now is the time to start aggregation because the month has changed. If the current time is not the time to start counting (NO in step S7), the processor 301 returns the process to step S1.

If the current time is the time to start tallying (YES in step S7), processor 301 performs tallying processing (step S8), and saves the tallying result in tally database 332 (step S9). Specifically, the processor 301 selects speakers (target speakers) one by one from the basic database 331 corresponding to meetings held during the target period (for example, the most recent month). The processor 301 then calculates an evaluation value for each previous speaker for the target speaker. The processor 301 adds a new record 332a corresponding to the pair of the target speaker and the previous speaker to the aggregation database 332 for each previous speaker. Since the description method for each field of the new record 332a is as described above, detailed description is omitted. After step S9, the processor 301 returns the process to step S1.

(Determination of pairs to be warned)
FIG. 13 is a flow chart showing the flow of processing for determining pairs to be warned. First, the processor 301 selects one target pair (step S11). For example, processor 301 selects a target pair consisting of a speaker and a previous speaker in response to input to input device 305A.

Next, the processor 301 reads data (record 332a) corresponding to the target pair from the summary database 332 (step S12). The processor 301 determines whether or not the compatibility tends to deteriorate based on the period described in the field 332d of the read record 332a and the evaluation values described in the fields 332e to 332h (step S13). Specifically, the processor 301 determines that the compatibility is deteriorating when the range of change in the evaluation value in the direction of worsening compatibility exceeds a reference value during a certain period of time in the past.

If the compatibility tends to deteriorate (YES in step S13), the processor 301 determines the target pair as a pair to be warned (step S14).

After step S14 and when the compatibility is not on a downward trend (NO in step S13), the processor 301 determines whether or not there is an unselected pair (step S15). If there is an unselected pair (YES in step S15), processor 301 returns the process to step S11. If there is no unselected pair (NO in step S15), processor 301 terminates the process.

(labeling process)
FIG. 14 is a flow chart showing the flow of labeling processing. First, the processor 301 selects one person to be labeled (step S21). For example, the processor 301 selects the labeling target according to the input to the input device 305A.

Next, the processor 301 reads data (record 332a) corresponding to the labeling pair from the total database 332 (step S22). Specifically, the processor 301 reads the record 332a in which the personal ID of the person to be labeled is described in the field 332b.

Based on the read record 332a, the processor 301 determines whether or not the number of newly acquainted people is large and whether or not there is a strong tendency for compatibility with newly acquainted people to increase in a short period of time (step S23). . Specifically, the processor 301 calculates the change width W1 of the evaluation value in the direction of improving compatibility in a certain period T1 after the start of compatibility evaluation for each pair whose speaker is the person to be evaluated. do. The processor 301 compares the representative value of the change width W1 for each pair with the first reference value. Furthermore, the processor 301 compares the number of pairs whose compatibility evaluation is started per unit time with the second reference value. The processor 301 performs step YES is determined in S23.

In the case of YES in step S23, the processor 301 attaches the label "good human relationship building" (first label) to the person to be labeled (step S24).

In the case of NO in step S23 and after step S24, the processor 301 determines based on the read record 332a whether or not the affinity with the newly acquainted person tends to increase over a long period of time ( step S25). Specifically, the processor 301 determines, for each pair whose speaker is the person to be labeled, the change in the evaluation value in the direction of improving compatibility during a certain period of time T2 (>T1) after the start of evaluation of compatibility. Calculate the width W2. The processor 301 compares the representative value of the change width W2 for each pair with the third reference value. The processor 301 determines YES in step S25 when the representative value of the change width W2 for each pair exceeds the third reference value.

If YES in step S25, the processor 301 labels the person to be labeled "positive" (second label) (step S26).

In the case of NO in step S25 and after step S26, the processor 301 determines whether or not the compatibility with others is generally low in the past fixed period based on the read record 332a (step S27). . Specifically, the processor 301 calculates a representative value of evaluation values for each pair in the past fixed period T3. The processor 301 determines YES in step S27 if the representative value is included in the target range R that is less compatible than the median value in the possible range of evaluation values.

In the case of YES in step S27, the processor 301 attaches the label "negative" (third label) to the person to be labeled (step S28).

In the case of NO in step S27 and after step S28, the processor 301 determines, based on the read record 332a, whether the compatibility with others has tended to deteriorate and the working hours have been long for a certain period of time in the past. is determined (step S29). Specifically, the processor 301 calculates the change width W3 of the evaluation value in the direction of worsening compatibility during the past fixed period T4 for each pair whose speaker is the person to be labeled. The processor 301 compares the representative value of the change width W3 for each pair with the fourth reference value. Furthermore, the processor 301 reads the working hours of the person to be labeled from the work database 333 and compares the read working hours with the fifth reference value. The processor 301 determines YES in step S29 when the representative value of the change width W3 for each pair exceeds the fourth reference value and the working hours exceed the fifth reference value.

In the case of YES in step S29, the processor 301 attaches the label "sign of job separation/illness" (fourth label) to the person to be labeled (step S30).

If NO in step S29 and after step S30, the processor 301 determines whether there is an unselected labeling target (step S31). If there is an unselected labeling target person (YES in step S31), the processor 301 returns the process to step S31. If there is no unselected labeling subject (NO in step S31), the processor 301 terminates the process.

<G. Variation>
(Modification 1)
The evaluation value calculated by the evaluation unit 34 is not limited to the types described above. For example, the evaluation unit 34 may calculate the evaluation value according to the following formula.
Evaluation value = a x P + b x Q + c x R
P indicates the ratio of utterances with high psychological safety. Q indicates the percentage of utterances with low psychological safety. R indicates total talk time. a, b, and c are predetermined coefficients. a and c are positive and b is negative. The evaluation value calculated according to the above formula increases as the compatibility increases.

Alternatively, the evaluation unit 34 may calculate a representative value (average value, median value, minimum value, etc.) of the psychological safety scores detected for the statements of each pair of evaluation subjects as an evaluation value for each period. good.

(Modification 2)
In the above description, the information processing apparatus 300 acquires voice data of the online conference as the statement by the speaker. However, the information processing apparatus 300 may acquire the speaker's speech by another method in addition to or instead of the voice data of the online conference.

For example, the information processing device 300 may acquire message data posted to a chat as a statement by a speaker. In this case, the compatibility between the target speaker and the contributor who posted the message immediately before the message by the target speaker (previous speaker) is evaluated.

Alternatively, the information processing device 300 may acquire e-mail data as the statement by the speaker. In this case, compatibility between the target speaker who sent a reply mail to a certain email and the sender of the email (previous speaker) is evaluated.

(Modification 3)
In the above description, it is assumed that the model generation unit 31 of the information processing device 300 generates the trained model 340 . However, trained model 340 may be generated by a device different from information processing device 300 and installed in information processing device 300 .

(Modification 4)
In the above description, the acquisition unit 32 , the detection unit 33 , the evaluation unit 34 , the provision unit 35 , the warning unit 36 and the labeling unit 37 are realized by the processor 301 of the information processing device 300 . However, some of the acquisition unit 32, the detection unit 33, the evaluation unit 34, the provision unit 35, the warning unit 36, and the labeling unit 37 are implemented by processors provided in one or more devices different from the information processing device 300. may In this case, the basic database 331 and the summary database may be stored in different devices.

It should be considered that the embodiments disclosed this time are illustrative in all respects and not restrictive. The scope of the present invention is indicated by the scope of the claims rather than the above description, and is intended to include all modifications within the scope and meaning equivalent to the scope of the claims.

1 system, 11,301 processor, 12,302 main memory, 13 camera, 14 microphone, 15,303 communication interface, 16,305 input interface, 16A, 305A input device, 17,306 display interface, 17A, 306A display, 18 speaker, 19,307 optical drive, 19A, 307A optical disk, 20,310 storage device, 22,320 program, 25,319 internal bus, 30 storage unit, 31 model generation unit, 32 acquisition unit, 33 detection unit, 34 evaluation Part 35 Providing Part 36 Warning Part 37 Labeling Part 50, 60 Screen 51 Area 52, 53, 61 Input Field 54 Warning Indicator 55 Label Display Field 62, 63 Display Field 100 Network 200 Terminal , 300 information processing device, 330 DB group, 331 basic database, 331a, 332a records, 331b to 331e, 332b to 332i fields, 332 summary database, 333 work database, 340 learned model.

Claims (14)

an acquisition unit that acquires a statement by a speaker;
a detection unit that detects the degree of psychological safety for each statement;
an evaluation unit that evaluates compatibility for each pair between the target speaker and another speaker who spoke immediately before the target speaker's utterance, based on the detected degree of the target speaker's utterance; Compatibility evaluation system. 3. The evaluation unit calculates an evaluation value of the compatibility for each pair by totaling the degree of detection of the utterances of the target speaker for each pair for each predetermined period. 1. The compatibility evaluation system according to 1. 3. The compatibility evaluation system according to claim 2, further comprising a providing unit that provides a screen showing changes in said evaluation values over time. The evaluation value includes a first numerical value indicating the number of times the degree of utterance exceeds the first standard, a second numerical value indicating the proportion of utterances with the degree exceeding the first standard, and the number of utterances with the degree falling below the second standard. Calculated using a third numerical value indicating the number of times, a fourth numerical value indicating the percentage of utterances whose degrees are below the second standard, and one or more numerical values selected from the first numerical value to the fourth numerical value. 4. A compatibility rating system according to claim 2 or 3, comprising at least one of the fifth numerical values. 2. For each of said pairs, further comprising a warning unit for issuing a warning indicating a sign of worsening compatibility in response to a change in said evaluation value in a direction of worsening compatibility exceeding a reference value within a certain period of time in the past. 5. The compatibility evaluation system according to any one of items 4 to 4. 5. The compatibility evaluation system according to any one of claims 2 to 4, further comprising a labeling unit that labels said target speaker based on the time change of said evaluation value for each pair. The labeling part is
For each pair, calculating a first change range of the evaluation value in the direction of improving compatibility in a first fixed period after the start of evaluation of compatibility,
when the first representative value of the first width of change for each pair exceeds a second reference value and the number of pairs for which the compatibility evaluation is started per unit time exceeds the first reference value 7. The compatibility evaluation system according to claim 6, wherein a first label is attached to said target speaker. The labeling part is
For each of the pair, calculating a second change range of the evaluation value in the direction of improving the compatibility in a second fixed period after the start of the compatibility evaluation,
8. The compatibility evaluation system according to claim 6, wherein a second label is given to said target speaker in accordance with a second representative value of said second change width for each pair exceeding a third reference value. . The labeling part is
calculating a third representative value of the evaluation values for each pair in the past certain period;
7. A third label is applied to the target speaker according to the fact that the third representative value is included in a target range that is less compatible than the median value of the possible range of the evaluation values. 9. The compatibility evaluation system according to any one of 8. The speaker is a worker,
The compatibility evaluation system further comprises a database that manages the working hours of each speaker,
The labeling part is
For each pair, calculating a third variation range of the evaluation value in the past fixed period in the direction of worsening compatibility,
When the fourth representative value of the third change width for each pair exceeds a fourth reference value and the working hours of the target speaker exceed a fifth reference value, 10. A compatibility evaluation system according to any one of claims 6 to 9, wherein a fourth label is applied. The detection unit acquires the degree by inputting data indicating the statement into a trained model,
The trained model is generated by machine learning using teacher data so as to output the degree of psychological safety of the person who made the statement when data indicating the statement is input. 11. The compatibility evaluation system according to any one of 1 to 10. 12. The compatibility according to any one of claims 1 to 11, wherein said acquisition unit acquires at least one of voice data, e-mail data, and message data posted in a chat as data indicating said statement. rating system. obtaining utterances by a speaker;
detecting a degree of psychological safety for each utterance;
Evaluating compatibility for each pair between the target speaker and another speaker who spoke immediately before the target speaker's utterance based on the degree detected for the target speaker's utterance. Method. A program for causing a computer to execute the method according to claim 13.

Images