Mental state analysis device, mental state analysis system, mental state analysis method and mental state analysis program

Abstract

To provide a mental state analysis device, a mental state analysis system, a mental state analysis method and a mental state analysis program capable of extracting voice data of a speaker who is an analysis object, from voice data of talk, and analyzing a mental state of the speaker on the basis of a result of emotion analysis and evaluating of the mental state of the speaker.SOLUTION: A mental state analysis device 1 comprises: an acquisition part acquiring voice data of talk; an extraction part for extracting voice data of a speaker who is an analysis object, from the acquired voice data; an emotion analysis part for analyzing the voice data of the speaker who is the analysis object for calculating a first score related to the emotion of the speaker; and a mental state analysis part for calculating a second score related to the mental state of the speaker on the basis of the first score.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a mind-body condition analyzing device, a mind-body condition analyzing system, a mind-body condition analyzing method, and a mind-body condition analyzing program. [Background technology]

[0002] Research is being conducted on emotion analysis methods that analyze a person's inner feelings based on the tone and pitch of their voice, facial expressions and gestures, written text and pictures, etc.

[0003] Technologies have been proposed that use this emotion analysis method to analyze dialogue voice data to support operators at contact centers who handle customer inquiries over the phone. For example, there are technologies shown in Patent Documents 1 and 2. These technologies analyze the customer's emotions (such as joy or anger) from the dialogue between the customer and the operator, and instruct the operator on how to respond accordingly. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2020-184243 A [Patent Document 2] JP 2020-115224 A [Patent Document 3] JP 2020-086818 A Summary of the Invention [Problem to be solved by the invention]

[0005] In the labor management of companies, it is important to understand the mental and physical state of workers, discover problems, and work to resolve them early. Although the understanding of mental and physical state is widely carried out by specialists such as industrial physicians, this method has the problem of being difficult to grasp the subject's mental and physical state on a regular basis.

[0006] As a technical method for grasping the mental and physical state, for example, a technology has been proposed for grasping the mental and physical state from an analysis of vital data acquired by a sensor attached to a subject, as shown in Patent Document 3. However, the subject needs to wear a sensor device, and even with this method, there remains the problem that it is difficult to grasp the subject's mental and physical state constantly.

[0007] On the other hand, in companies, telephone conversations are routinely used as a means of internal and external communication, and it is easy to regularly obtain the voice data of subjects. Therefore, if the mental and physical state of people can be understood and analyzed using voice data, the aforementioned issues can be resolved. In addition, with the recent spread of remote work, the use of voice data is also effective in non-face-to-face labor management.

[0008] However, the aforementioned technology for supporting contact center operators by analyzing emotions in voice data has a problem in that the technology cannot be directly adapted for other uses because the purpose is different.

[0009] Specifically, while operator support requires immediate analysis, analysis based on voice data over a certain period is necessary for mental and physical state analysis. Also, for mental and physical state analysis, it is necessary to extract the voice data of the speaker to be analyzed from a large amount of voice data over a certain period. Furthermore, for mental and physical state analysis, it is necessary to evaluate and derive the mental and physical state (positive or negative attitude toward work, etc.) from the emotions of the subject (happiness, anger, etc.) obtained through emotion analysis.

[0010] The present invention aims to provide a mind-somatic condition analysis device, a mind-somatic condition analysis system, a mind-somatic condition analysis method, and a mind-somatic condition analysis program that extracts voice data of a speaker to be analyzed from the voice data of a conversation, and further analyzes and evaluates the speaker's mental and physical condition based on the results of emotion analysis. [Means for solving the problem]

[0011] One aspect of the present invention is a mental and physical state analysis device that is characterized by comprising an acquisition unit that acquires voice data of a conversation, an extraction unit that extracts voice data of a speaker to be analyzed from the acquired voice data, an emotion analysis unit that analyzes the extracted voice data of the speaker to be analyzed and calculates a score for the speaker's emotions, and a mental and physical state analysis unit that calculates a score for the speaker's mental and physical state from the score for the speaker's emotions.

[0012] One aspect of the present invention is a mental and physical state analysis device, which is characterized by further comprising an evaluation unit that evaluates the speaker's mental and physical state based on a score for the speaker's mental and physical state, and an output unit that outputs the result of the evaluation by the evaluation unit.

[0013] A mind-body state analyzing device according to one aspect of the present invention is characterized in that an extracting unit extracts voice data of a speaker to be analyzed using a code for identifying the speaker that is associated with the voice data.

[0014] A mental and physical state analyzing device according to one aspect of the present invention is characterized in that a mental and physical state analyzing section calculates scores of two indices, positive and negative, regarding the speaker's mental and physical state.

[0015] A mind-body state analyzing device according to one aspect of the present invention is characterized in that an emotion analyzing unit calculates scores for each of the indices of normality, joy, anger, sadness, and vitality.

[0016] One aspect of the mental and physical state analysis device of the present invention is characterized in that the mental and physical state analysis unit weights the scores calculated by the emotion analysis unit for each of the indices of normality, joy, anger, sadness, and vitality, and calculates scores for two indices, positive and negative, for the speaker's mental and physical state.

[0017] One aspect of the present invention is a mental and physical state analysis system that comprises a mental and physical state analysis device, a communications network for transmitting voice data, and nodes connected to the communications network, and is characterized in that an acquisition unit of the mental and physical state analysis device acquires the voice data exchanged between the nodes via the communications network.

[0018] A mental and physical state analysis method according to one aspect of the present invention is characterized by comprising an acquisition step of acquiring voice data of a conversation, an extraction step of extracting voice data of a speaker to be analyzed from the acquired voice data, an emotion analysis step of analyzing the extracted voice data to calculate a score for the speaker's emotion, and a mental and physical state analysis step of calculating a score for the speaker's emotion or a score for the speaker's mental and physical state.

[0019] One aspect of the present invention is a mental and physical state analysis program that causes a computer to execute the following steps: an acquisition step of acquiring voice data of a conversation; an extraction step of extracting voice data of a speaker to be analyzed from the acquired voice data; an emotion analysis step of analyzing the extracted voice data to calculate a score for the speaker's emotions; and a mental and physical state analysis step of calculating a score for the speaker's mental and physical state from the score for the speaker's emotions. Effect of the Invention

[0020] The mental and physical state analysis device, mental and physical state analysis system, mental and physical state analysis method, and mental and physical state analysis program according to the present invention extract voice data of the speaker to be analyzed from the voice data of the conversation, and further analyze and evaluate the speaker's mental and physical state based on the results of emotion analysis. [Brief description of the drawings]

[0021] [Figure 1] FIG. 2 is a block diagram of a mind-body state analyzing device. [Diagram 2] FIG. 2 is a hardware configuration diagram of the mind and body state analyzing device. [Diagram 3] FIG. 1 is a configuration diagram of a mind and body state analysis system. [Figure 4] 4 is a flowchart showing the operation of the mind-body state analyzing device. [Diagram 5] 13 is a flowchart of a mental and physical state analysis process. [Figure 6]13 is a diagram for explaining calculation of a vitality score element of an index indicating a physical and mental state from vitality score elements. FIG. [Figure 7] 13 is a graph showing the relationship between the vitality score and the percentage of the cumulative frequency of the number of data items. [Figure 8] This figure explains the conversion from values ​​that set areas based on the relationship between the vitality score and the cumulative frequency ratio of the number of data to values ​​that set areas for the speaker's mental and physical state using positive scores. [Figure 9] 13 is an example of a report that is output. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0022] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to the drawings. Duplicate descriptions will be omitted, and the same or corresponding parts in each drawing will be given the same reference numerals.

[0023] In an embodiment of the present invention, the mental and physical state analysis device, mental and physical state analysis system, mental and physical state analysis method, and mental and physical state analysis program extract voice data of the speaker to be analyzed from voice data of a conversation acquired from outside, and based on the results of the emotion analysis, further analyze and evaluate the speaker's mental and physical state, and output the results. EXAMPLES

[0024] First, the mental and physical state analyzer will be described. Fig. 1 is a block diagram of the mental and physical state analyzer. The mental and physical state analyzer 1 may be configured as a standalone device, or may be incorporated into other devices for use. The other devices incorporating the mental and physical state analyzer 1 may be, for example, the main body of the mental and physical state analyzer, as well as communication network components such as a business voice switch, a PBX (private branch exchange), and a router, or electrical appliances such as a personal computer, a smartphone, and a portable information terminal.

[0025] Fig. 2 is a hardware configuration diagram of the mind-body state analyzing device 1. As shown in Fig. 2, physically, it is configured as a computer including a central processing unit (CPU) 201, an input device 202, an output device 203, a main storage device (RAM / ROM) 204, and an auxiliary storage device 205.

[0026] Each function of the mind-body state analysis device 1 is realized by loading a mind-body state analysis program into a main memory device (RAM / ROM) 204 etc. shown in Figure 2, operating an input device 202 and an output device 203 under the control of a central processing unit (CPU) 201, and reading and writing data to and from the main memory device (RAM / ROM) 204 and the auxiliary memory device 205.

[0027] 1, the mental and physical state analysis device 1 includes an acquisition unit 101, an extraction unit 102, a feeling analysis unit 103, a mental and physical state analysis unit 104, an evaluation unit 105, and an output unit 106. Dialogue voice data is provided from the outside to the mental and physical state analysis device 1 via the acquisition unit 101. In addition, the mental and physical state analysis device 1 outputs information to the outside via the output unit 106.

[0028] The function of each block of the mind-body state analyzer 1 will be described according to the block diagram of Figure 1. The detailed operation of each block will be described later.

[0029] The acquisition unit 101 acquires voice data of a conversation from an external source.

[0030] The extraction unit 102 extracts speech data of a speaker to be analyzed from the speech data acquired by the acquisition unit 101 .

[0031] The emotion analysis unit 103 analyzes the voice data extracted by the extraction unit 102 and calculates a score for the emotion of the speaker. The score for the emotion of the speaker will be described later.

[0032] The mental and physical state analysis unit 104 calculates a score for the mental and physical state of the speaker from the score for the emotion of the speaker calculated by the emotion analysis unit 103. The score for the mental and physical state of the speaker will be described later.

[0033] The evaluation unit 105 evaluates the mental and physical state of the speaker based on the score for the mental and physical state of the speaker calculated by the mental and physical state analysis unit 104. A specific method of evaluation will be described later.

[0034] The output unit 106 outputs the evaluation result of the mental and physical state of the speaker evaluated by the evaluation unit 105. The output evaluation result will be described later in detail.

[0035] Next, the mental and physical state analysis system will be described. Fig. 3 is a diagram showing the configuration of a mental and physical state analysis system having a mental and physical state analysis device 1 as a component. The mental and physical state analysis device 1 functions to achieve a purpose by itself, but also functions to achieve a purpose as a mental and physical state analysis system in which the mental and physical state analysis device 1 is connected to a communication network 301 together with one or more nodes 302 as shown in Fig. 3.

[0036] Here, the communication network 301 refers to a mechanism for communication between connected devices, such as a LAN (local area network), a WAN (wide area network), or the Internet, regardless of the method, standard, or whether it is wired or wireless.

[0037] Here, the node 302 refers to a device that is connected to the communication network 301 and can communicate via the communication network 301. For example, it corresponds to a business voice exchange or PBX (Private Branch eXchange) that processes the voice data of the conversation, an external storage device that accumulates the voice data of the conversation, a personal computer, a smartphone, a portable information terminal, and the like that function as an external input / output device of the mind-body state analysis device 1. Note that these are merely examples, and the node 302 is not limited to these.

[0038] If node 302 is a device that processes voice data, such as a PBX, then mental and physical state analysis device 1 can acquire voice data via communication network 301. If node 302 is a personal computer or the like that functions as an external input / output device, node 302 can input voice data stored outside the mental and physical state analysis system to mental and physical state analysis device 1 via communication network 301, and output the results of evaluation of the speaker's mental and physical state to node 302.

[0039] Next, the operation of the mind-body state analyzer 1 according to this embodiment will be described. FIG. 4 is a flowchart showing the operation of the mind-body state analyzer 1 according to this embodiment. The flowchart in FIG. 4 also shows the steps of the mind-body state analyzing method according to this embodiment. The mind-body state analyzer 1 executes the steps of the mind-body state analyzing method according to this embodiment by operating according to the flowchart in FIG. 4. Below, the operation of the mind-body state analyzer 1 according to this embodiment will be described according to the flowchart in FIG. 4.

[0040] The operation of the mind-body state analyzer 1 may be started automatically, for example, after input of voice data, or may be started by an explicit command. When the mind-body state analyzer 1 starts its operation, it starts the process of the flowchart in FIG.

[0041] Each process shown in the flowchart of FIG. 4 may be started automatically after the pre-processing is completed, or may be started by an explicit command to start the process.

[0042] 4 starts, the acquisition unit 101 executes an acquisition process (S401) in the mind-body state analyzing device 1. In the acquisition process (S401), voice data of a conversation is acquired from the outside.

[0043] Here, the conversational voice data refers to voice data of a conversation between multiple people over the telephone or the like, and may be one-to-one, one-to-many, or many-to-many. The data format of the voice data is not important.

[0044] In this embodiment, it is assumed that a code for identifying a speaker is attached to the voice data of the dialogue. Here, the code refers to additional information for identifying the speaker of the voice data of the dialogue. For example, it corresponds to a telephone number assigned to a telephone terminal used by the speaker, or an IP address assigned to an information terminal used by the speaker. The code for identifying the speaker is used in the extraction process (S402) described later. It is also possible to perform the extraction process (S402) without using a code for identifying a speaker by using a voice recognition technology or the like. In this embodiment, it is assumed that a code for identifying a speaker is attached to the voice data of the dialogue in consideration of ease of understanding of the explanation, and it is not essential for the present invention that a code for identifying a speaker is attached to the voice data of the dialogue.

[0045] In the following of this embodiment, in consideration of ease of understanding of the explanation, an example will be given in which the voice data of the dialogue uses a telephone number assigned to the telephone terminal used by the speaker as a code for identifying the speaker.

[0046] The acquisition process (S401) may be performed sequentially, or voice data for a certain period may be acquired collectively. An example of performing the acquisition process (S401) sequentially is connecting the mind-body state analysis device 1 to a device that processes dialogue voice data (e.g., a PBX) to constantly acquire voice data. An example of performing the acquisition process (S401) collectively is storing voice data for a certain period acquired via a device that processes dialogue voice data (e.g., a PBX) and importing the voice data into the mind-body state analysis device 1 in a batch process.

[0047] After the acquisition process (S401) is completed, the extraction unit 102 of the mind-body state analysis device 1 next executes the extraction process (S402). In the extraction process (S402), voice data of the speaker to be analyzed is extracted from the acquired voice data. As described above, in this embodiment, it is assumed that the voice data of the dialogue is accompanied by a code for identifying the speaker, and a telephone number assigned to the telephone terminal used by the speaker is used as the code for identifying the speaker. In the extraction process (S402), voice data linked to the telephone number assigned to the telephone terminal used by the speaker is extracted as the voice data of the speaker to be analyzed.

[0048] As described above, in addition to this embodiment, it is also possible to perform the extraction process (S402) without using a code for identifying the speaker by using voice recognition technology, etc. In such a case, it is not necessary to include a code for identifying the speaker.

[0049] In this embodiment, the voice data of the speaker to be analyzed that is extracted in the extraction process (S402) is extracted as voice data for each call. Here, a call refers to one communication from connection to disconnection. In addition, voice data is extracted for calls in which the speaker to be analyzed is the caller or the callee.

[0050] In the case where the voice data acquired in the acquisition process (S401) is accompanied by information on the date and time of the voice conversation, such as the date and time of connection and disconnection of communication, as in telephone conversation data, in this embodiment, the extraction process (S402) extracts voice data for a specified period by specifying a start date and time and an end date and time, etc. By extracting voice data for a specified period, it is possible to analyze the mental and physical state for the specified period and reduce the amount of data to be processed.

[0051] After the extraction process (S402) is completed, the emotion analysis unit 103 of the mind-body state analysis device 1 executes emotion analysis process (S403). In the emotion analysis process (S403), the voice data of the extracted speaker to be analyzed is analyzed to calculate a score for the speaker's emotion.

[0052] In this embodiment, the score regarding the emotion of the speaker calculated in the emotion analysis process (S403) is calculated for each call.

[0053] Here, the score for the speaker's emotion refers to a value obtained by quantifying the emotion of the speaker when he / she speaks, using multiple indicators of emotion through emotion analysis processing (S403). In this embodiment, the indicators of emotion are five items: normal, joy, anger, sadness, and vitality. In this embodiment, vitality is an indicator of the degree of energy that is the source of mental and physical activity, such as human energy, willpower, activity, and motivation, which is required for humans to be generally active, and is defined as a relative degree from each person's subjective point of view.

[0054] The method of emotion analysis processing (S403) may be any method that can analyze the extracted voice data of the speaker to be analyzed and calculate a quantitative score for the speaker's emotion. For example, Patent Document 1 discloses a method of analyzing the emotion of a speaker by comparing the voice data of the speaker to be analyzed with a model that has learned acoustic characteristics (voice pitch, speaking speed, loudness, intonation, etc.) for each emotion such as neutral, joy, and anger.

[0055] In addition, the method of comparing the model that has learned the acoustic characteristics for each emotion with the speech data of the speaker to be analyzed is not important as long as it is possible to calculate the similarity between them as a quantitative score. Known methods include, for example, calculating the similarity as a quantitative score using Euclidean distance, Pearson's probability correlation coefficient, Jaccard coefficient, etc.

[0056] In this embodiment, the following description will be given on the assumption that the indices indicating emotions are five items: normal, joy, anger, sadness, and vitality, and the scores for each index are quantified within a range from a minimum value of 0 to a maximum value of 20. Note that the process of quantifying each of the indices indicating the five emotions, normal, joy, anger, sadness, and vitality, within a range from a minimum value of 0 to a maximum value of 20 shown in this embodiment is merely an example, and other indices that appropriately indicate the speaker's emotions may be adopted in the present invention, and the quantification is not limited to a range from a minimum value of 0 to a maximum value of 20.

[0057] When the emotion analysis process (S403) is completed, the mental and physical state analysis unit 104 of the mental and physical state analysis device 1 then executes the mental and physical state analysis process (S404). In the mental and physical state analysis process (S404), a score for the speaker's mental and physical state is calculated from the score for the speaker's emotion calculated in the emotion analysis process (S403).

[0058] Here, the mental and physical state refers to the mental and physical state a person is in when performing their duties. Examples of mental states include being full of motivation for a new job or being worried about interpersonal relationships at work, while an example of a physical state is being in poor health and causing problems at work.

[0059] In this embodiment, the score for the speaker's mental and physical state is a value quantitatively indicated for two indicators, positive and negative. This is to provide an effective judgment indicator for determining whether the speaker is in a mental and physical state where he or she is proactive about work or in a mental and physical state where he or she is reluctant about work and has problems in the labor management of a company, etc., so the score for the two indicators, positive and negative, is always 100. In this embodiment, the score for the speaker's mental and physical state is set to two indicators, positive and negative, and the total score is always 100, in addition to being effective in the labor management of a company, etc. according to the present invention, it is also in consideration of the ease of explanation and understanding. It is not prohibited to set three or more indicators for the speaker's mental and physical state in the present invention, and to set the total score for the two indicators, positive and negative, to another value.

[0060] In the following, in this embodiment, as described above, it is assumed that a quantified score is calculated for each call in the emotion analysis process (S403) for each index indicating five emotions, namely, normal, joy, anger, sadness, and vitality, with a minimum value of 0 and a maximum value of 20. Specifically, for example, the score for one call associated with a speaker can be expressed as Example A (normal, joy, anger, sadness, vitality) = (8, 6, 15, 12, 9) or Example B (normal, joy, anger, sadness, vitality) = (11, 15, 7, 8, 12). In this embodiment, the larger the score value, the greater the degree of emotion for each index. Specifically, for example, the larger the anger score, the greater the degree of the speaker's anger.

[0061] 5 is a flowchart of the mental and physical state analysis process (S404). Let us now leave FIG. 4 and explain the mental and physical state analysis process (S404) according to the flowchart of FIG.

[0062] When the mental and physical state analysis process (S404) starts, the score calculated in the emotion analysis process (S403) is obtained (S501). As described above, the score obtained in the process of S501 is a quantified score for each call with a minimum value of 0 and a maximum value of 20 for each index showing the five emotions of normal, joy, anger, sadness, and vitality.

[0063] After the process of acquiring a score indicating emotion (S501) is completed, the process of dividing each element of the score indicating emotion into two categories, positive and negative, starts (S502). The process of S502 is executed for each score for each call. This process divides the score based on the index indicating the five emotions of neutral, joy, anger, sadness, and vitality calculated in the emotion analysis process (S403) into an index indicating the mental and physical state (positive / negative).

[0064] In the process of S502 in this embodiment, the elements of the score for each call for each of the five emotion indicators are sorted into indicators indicating mental and physical states (positive / negative) based on the following rules. Rule 1: The elements of the normal and joyful scores are elements of the positive score. Rule 2: The score components of anger and sadness are included in the negative score components. Rule 3: The standard value for dividing the vitality score elements into positive and negative is 10. If the value is greater than 10, the score element is calculated as positive, and if the value is less than 10, the score element is calculated as negative.

[0065] Fig. 6 is a diagram for explaining the calculation of vitality score elements of an index (positive / negative) indicating a physical and mental state from the vitality score elements shown as rule 3 above. Here, the vitality score element is G, the positive vitality score element in the index indicating a physical and mental state is Gp, and the negative vitality score element is Gn. The same symbols are used in Fig. 6 as well.

[0066] The processing of rule 3 will be described with reference to Figure 6. Element G of the vitality score is obtained (S601). In the process of S602, it is determined whether element G of the vitality score is greater than 10. If G>10, Gp=(G-10) is calculated as an element of the positive vitality score (S603). Next, in the process of S604, it is determined whether element G of the vitality score is less than 10. If G<10, Gn=-(G-10) is calculated as an element of the negative vitality score (S605).

[0067] When the process of S502 is applied to the example score A (normal, happy, anger, sadness, vitality) = (8, 6, 15, 12, 9) based on the index indicating the five emotions given as specific examples above, it is assigned as follows: Positive: (normal, happy, energetic) = (8, 6, NULL) Negative: (anger, sadness, energy) = (15, 12, 1) In addition, the elements of the score of positive vitality are NULL and have no value. Therefore, NULL elements are treated as not existing data in the subsequent processing.

[0068] Similarly, when the process of S502 is applied to the example score B (normal, happy, anger, sadness, vitality) = (11, 15, 7, 8, 12) based on the index indicating the five emotions given as specific examples above, it is assigned as follows: Positive: (normal, happy, energetic) = (11, 15, 2) Negative: (anger, sadness, vitality) = (7, 8, NULL) In addition, the elements of the score of the negative vitality level are NULL and have no value. Therefore, NULL elements are treated as not existing data in the subsequent processing.

[0069] When the process of S502 ends, the process of weighting the score (S503) starts. The process of S503 is executed for each score for each call. This is a process of making corrections according to the environment so that the score for the speaker's mental and physical condition, calculated as a quantitative value for two indicators, positive and negative, in the mental and physical condition analysis process (S404) becomes a more effective judgment indicator for labor management in companies, etc.

[0070] In the process of S503, a weighting coefficient for each index indicating the five emotions, namely, normal, joy, anger, sadness, and vitality, is multiplied by each score element. If the weighting coefficients corresponding to the indexes indicating the five emotions, namely, normal, joy, anger, sadness, and vitality, are a, b, c, d, and e, respectively, the elements allocated to the index (positive / negative) indicating the mental and physical state corresponding to the score example based on the index indicating the five emotions given as a specific example above will be as follows by the process of S503.

[0071] Elements classified as indicators (positive / negative) showing mental and physical state corresponding to score example A Positive: (normal, happy, energetic) = (8a, 6b, NULL) Negative: (anger, sadness, energy) = (15c, 12d, 1e)

[0072] Elements classified as indicators (positive / negative) showing mental and physical state corresponding to score example B Positive: (normal, happy, energetic) = (11a, 15b, 2e) Negative: (anger, sadness, vitality) = (7c, 8d, NULL)

[0073] In this embodiment, a = 0.19, b = 0.21, c = 0.25, d = 0.15, and e = 0.20. This is a set of weighting coefficients so that when the average score based on each index showing the five emotions is calculated from 807 sample data of calls lasting 10 seconds or more and then the average score is divided into indexes showing mental and physical states (positive / negative), the ratio of positive to negative is 1:1.

[0074] The weighting coefficients and the calculation method thereof described above are merely examples, and the present invention is not limited thereto. In the environment of the company or the like in which the present invention is implemented, corrections may be made according to the environment to make the index more effective. For example, in the environment in which 807 sample data of calls longer than 10 seconds are obtained, it has been found that an evaluation more suited to the environment can be obtained by correcting the ratio of positive to negative to 51.5:48.5.

[0075] When the process of S503 ends, the process of calculating the score based on the index (positive / negative) indicating the mental and physical state (S504) starts. As described above, in this embodiment, the sum of the scores for the two indexes, positive and negative, is always 100.

[0076] In the process of S504, first, the elements for each call that were classified as positive and negative in the process of S503 are extracted and added. As an example, if the above-mentioned score example A and score example B are all the data, it is as follows. Positive: (8a+6b)+(11a+15b+2e) Negative: (15c+12d+1e)+(7c+8d)

[0077] In the process of S504, the elements assigned to the two indicators, positive and negative, are corrected so that the total value is 100, and the positive and negative scores are calculated. As an example, if the above-mentioned score example A and score example B are all the data, it is as follows. Positive score = 100 x {(8a + 6b) + (11a + 15b + 2e)} / {(8a + 6b) + (11a + 15b + 2e) + (15c + 12d + 1e) + (7c + 8d)} Negative score = 100 x {(15c + 12d + 1e) + (7c + 8d)} / {(8a + 6b) + (11a + 15b + 2e) + (15c + 12d + 1e) + (7c + 8d)}

[0078] When the process of S504 completes the calculation of the score based on the index (positive / negative) indicating the mental and physical state, the mental and physical state analysis process (S404) ends.

[0079] Returning to Fig. 4, the explanation of the process will continue. When the mental and physical state analysis process (S404) is completed, the evaluation unit 105 of the mental and physical state analyzer 1 then executes the evaluation process (S405). In the evaluation process (S405), the mental and physical state of the speaker is evaluated from the scores for the two indexes, positive and negative, calculated in the mental and physical state analysis process (S404).

[0080] In this embodiment, a five-level evaluation is set, from the most negative level 1 to the most positive level 5, and an evaluation process (S405) is described in which an evaluation is made based on the scores for the two positive and negative indicators calculated in the mental and physical condition analysis process (S404).

[0081] First, the setting of the evaluation stages will be described. In this embodiment, the five stages from level 1 to level 5 are set using the vitality level, which affects both the positive and negative indicators.

[0082] FIG. 7 is a diagram for explaining the level setting for evaluation, and is a graph showing the relationship between the vitality score and the percentage of the cumulative frequency of the number of data. In the graph of FIG. 7, the horizontal axis is the vitality score, which has a range of values ​​from 0 as a minimum to 20 as a maximum. The vertical axis shows the percentage of the cumulative frequency of the number of data. The data is data from the environment of a company or the like in which the present invention is implemented, and is a sufficient number of data to set a level for evaluation. As shown in FIG. 7, P1 to P5, which show the percentage of the cumulative frequency of the number of data, respectively correspond to the vitality score divided into five equal parts (4, 8, 12, 16, 20). P1 to P5, which show the percentage of the cumulative frequency of the number of data, are values ​​that set a five-level range from level 1 to level 5.

[0083] As part of the evaluation process (S405), the evaluation unit 105 calculates values ​​P1 to P5 that set a five-level range from level 1 to level 5 from a sufficient number of data items that are data items in the environment of a company or the like in which the present invention is implemented and that are sufficient to set levels for evaluation.

[0084] The mental and physical state analyzing device 1 of the present invention may be configured to hold values ​​P1 to P5 that set a five-level range from level 1 to level 5 calculated in advance from a sufficient amount of data in the data of the environment of a company or the like implementing the present invention to set levels for evaluation. The values ​​P1 to P5 calculated from the above-mentioned 807 sample data of calls of 10 seconds or more were P1=0.07, P2=0.37, P3=0.77, P4=0.97, and P5=1.00. This refers to a configuration in which the mental and physical state analyzing device 1 holds these values.

[0085] Next, in the evaluation process (S405), the speaker's mental and physical state is evaluated into five levels from level 1 to level 5. The speaker's mental and physical state is evaluated using a positive score. For this purpose, the following conversion is performed from values ​​(P1 to P5) that set the range of five levels from level 1 to level 5, which are set based on the relationship between the vitality score and the cumulative frequency ratio of the number of data items described above.

[0086] FIG. 8 is a diagram illustrating the conversion from values ​​(P1 to P5) that set a five-level range from level 1 to level 5, which are set based on the relationship between the vitality score and the cumulative frequency ratio of the number of data, to values ​​(Q1 to Q5) that set a five-level range from level 1 to level 5 for the speaker's mental and physical state using positive scores.

[0087] For the same data as that for which the relationship between the vitality score and the cumulative frequency ratio of the number of data items was calculated as explained in Fig. 7, Fig. 8 shows the cumulative frequency distribution of the number of data items with the positive score on the horizontal axis. The values ​​(P1 to P5) that set the five-level range from level 1 to level 5 on the vertical axis are the same in Fig. 7 and Fig. 8.

[0088] In Figure 8, Q1 to Q5 on the horizontal axis are positive scores that correspond to the values ​​(P1 to P5) that set the five-level range from level 1 to level 5 on the vertical axis, and Q1 to Q5 are values ​​that set the five-level range from level 1 to level 5 for positive scores.

[0089] As an evaluation process (S405), the evaluation unit 105 calculates values ​​Q1 to Q5 that set a five-level range from level 1 to level 5 for positive scores, via P1 to P5, from data in the environment of a company or the like in which the present invention is implemented, which is a sufficient number of data to set levels for evaluation.

[0090] The mental and physical state analyzing device 1 of the present invention may be configured to hold values ​​Q1 to Q5 that set a five-level range from level 1 to level 5 calculated in advance from a sufficient number of data items that are data items in the environment of a company or the like implementing the present invention and that are sufficient to set levels for evaluation. The values ​​Q1 to Q5 calculated from the above-mentioned 807 sample data items of calls lasting 10 seconds or more were Q1=45, Q2=48, Q3=55, Q4=58, and Q5=100 (theoretical values). This refers to a configuration in which the mental and physical state analyzing device 1 holds these values.

[0091] In the evaluation process (S405), the speaker's emotional score and mental and physical state are evaluated on a five-level scale from level 1 to level 5 using values ​​Q1 to Q5 that set a five-level range from level 1 to level 5 for the calculated positive score. Level 1: Positive score less than Q1 Level 2: Positive scores are greater than or equal to Q1 and less than Q2 Level 3: Positive scores are between Q2 and Q3 Level 4: Positive scores are greater than or equal to Q3 and less than Q4 Level 5: Positive score of Q4 or more

[0092] When the evaluation process (S405) is completed, the output unit 106 of the mental and physical state analysis device 1 next executes the output process (S406). In the output process (S406), the evaluation result in the evaluation process (S405) is output. In this embodiment, in addition to this, the output unit 106 outputs the score on the emotion of the speaker calculated in the emotion analysis process (S403) and the score on the mental and physical state of the speaker calculated in the mental and physical state analysis process (S404) by the output process (S406). In this embodiment, the output unit 106 also outputs the evaluation result and each score by the output process (S406) in a report format that can support grasping the mental and physical state of workers, etc., discovering problems, and working to solve them early, which is an important issue in labor management of companies, etc. Furthermore, in this embodiment, the output unit 106 generates and outputs comments such as recommended actions to be taken according to the evaluation results in the output process (S406).

[0093] 9 is an example of a report output by the output process (S406). The report is output from the output unit 106 to an external output device. For example, the report may be displayed on a computer screen or printed on paper by a printer.

[0094] In Fig. 9, 901 is a code for identifying the speaker. In Fig. 9, it is shown as (TEST1). In this embodiment, it may be a code for identifying the speaker that is attached to the voice data of the dialogue, or it may be displayed by referring to related information linked thereto. For example, 901 may be a telephone number assigned to the speaker's telephone terminal, the speaker's name, etc.

[0095] In Fig. 9, 902 indicates the period during which the call data was analyzed. In this embodiment, when the voice data acquired in the acquisition process (S401) is accompanied by date and time information on the date and time of the voice conversation, such as the date and time of connection and disconnection of the communication, as in the case of telephone call data, an extraction process (S402) is also performed to extract the voice data for the period with the specified start date and end date and time. The specified period is displayed in 902.

[0096] In Fig. 9, 903 is a score regarding the emotion of the speaker calculated in the emotion analysis process (S403). In Fig. 9, it is displayed as a radar chart. This is an example, and it may be a numerical table or other graphs.

[0097] In Fig. 9, 904 denotes the score for the speaker's mental and physical state calculated in the mental and physical state analysis process (S404). In Fig. 9, it is displayed as a bar graph. This is an example, and it may be a numerical table or other graphs.

[0098] 9, 905 denotes the evaluation result in the evaluation process (S405). In addition to displaying the level of the evaluation result, in this embodiment, the recommended actions to be taken, etc., generated in the evaluation process (S405) are displayed as comments.

[0099] The above is a description of the mind and body state analyzer.

[0100] Next, a calculation program for causing a computer to function as a mind-body state analyzer will be described. The configuration of the computer is as shown in FIG.

[0101] The calculation program includes a main module, an input / output module, and a calculation processing module. The main module is a part that controls the overall processing. The input / output module includes an acquisition module and an output module. The calculation processing module includes an extraction module, an emotion analysis module, a mental and physical state analysis module, and an evaluation module. Functions realized by executing the main module, the input / output module, and the calculation processing module are similar to the functions of the acquisition unit 101, the extraction unit 102, the emotion analysis unit 103, the mental and physical state analysis unit 104, the evaluation unit 105, and the output unit 106 of the mental and physical state analysis device 1, respectively.

[0102] The calculation program may be provided by, for example, a storage medium such as a ROM or a semiconductor memory. The calculation program may also be provided via a network.

[0103] The above is a description of the embodiment.

[0104] As described above, according to the present application, the mental and physical state analysis device, mental and physical state analysis system, mental and physical state analysis method, and mental and physical state analysis program can extract the voice data of the speaker to be analyzed from the voice data of the dialogue, and further analyze and evaluate the speaker's mental and physical state based on the result of emotion analysis. This can support understanding the mental and physical state of workers, etc., discovering problems, and working to solve them early, which is an important issue in the labor management of companies, etc. [Explanation of symbols]

[0105] 1. Psychological and physical condition analyzer 101 Acquisition Department 102 Extraction part 103 Emotion Analysis Department 104 Mental and Physical Condition Analysis Department 105 Evaluation Department 106 Output section 201 Central Processing Unit (CPU) 202 Input Device 203 Output Device 204 Main memory (RAM / ROM) 205 Auxiliary storage device 301 Communication Network 302 Node 901 Speaker Identification Code 902 Period during which call data was analyzed 903 Speaker Emotion Score 904 Speaker's mental and physical state score 905 Evaluation result (level)

Claims

1. An acquisition unit for acquiring voice data of a dialogue; an extraction unit that extracts speech data of a speaker to be analyzed from the acquired speech data; a sentiment analysis unit that analyzes the extracted speech data of the speaker to be analyzed and calculates a first score regarding the sentiment of the speaker; a mental and physical state analysis unit that calculates a second score for the speaker's mental and physical state from the first score; A mental and physical state analyzing device comprising:

2. an evaluation unit that evaluates a mental and physical state of the speaker based on the second score; an output unit that outputs a result of the evaluation by the evaluation unit; 2. The mind and body state analyzing device according to claim 1, further comprising:

3. The extraction unit is extracting the speech data of the speaker to be analyzed based on a code for identifying the speaker that is associated with the speech data; 3. The mind and body state analyzing apparatus according to claim 1 or 2.

4. The mental and physical state analysis unit is Calculating a second score for two indicators, positive and negative, regarding the speaker's mental and physical state; 4. The mind and body state analyzing device according to claim 3.

5. The emotion analysis unit is Calculate a first score for each of the indices of normality, joy, anger, sadness, and vitality; 5. The mind and body state analyzing device according to claim 4.

6. The mental and physical state analysis unit is The first score is weighted with respect to each of the indices of normality, joy, anger, sadness, and vitality to calculate the second score.

6. The mind and body state analyzing device according to claim 5.

7. The mind and body state analyzer according to claim 1 or 2, a network for transmitting the audio data; a node connected to the network; Equipped with The acquisition unit acquires the voice data exchanged between the nodes via the network. A mental and physical state analysis system.

8. An acquisition step of acquiring voice data of the dialogue; An extraction step of extracting speech data of a speaker to be analyzed from the acquired speech data; a sentiment analysis step of analyzing the extracted speech data of the speaker to be analyzed and calculating a first score regarding the sentiment of the speaker; a mental and physical state analysis step of calculating a second score for the speaker's mental and physical state from the first score; A mental and physical state analysis method comprising:

9. An acquisition step of acquiring voice data of the dialogue; An extraction step of extracting speech data of a speaker to be analyzed from the acquired speech data; a sentiment analysis step of analyzing the extracted speech data of the speaker to be analyzed and calculating a first score regarding the sentiment of the speaker; a mental and physical state analysis step of calculating a second score for the speaker's mental and physical state from the first score; A mental and physical state analysis program characterized by causing a computer to execute the above.

Images