JP7238940B2 - Information processing device, information processing method and information processing program - Google Patents

Description

The present invention relates to an information processing device, an information processing method, and an information processing program.

In the above technical field, Patent Document 1 discloses that customer information, customer behavior history information, and a customer extraction rule for considering customers with high cancellation risk are compared, and customers who satisfy the customer extraction rule have a high cancellation risk. A technology for extracting as a customer is disclosed. In addition, Non-Patent Document 1 discloses a method of detecting a speech period. Non-Patent Document 2 discloses a method of emotion recognition. Non-Patent Document 3 discloses a machine learning method. Non-Patent Document 4 discloses a method of detecting a speech period.

Japanese Patent Application Laid-Open No. 2002-334200

Yusuke Kida and Tatsuya Kawahara, "Voice Activity Detection based on Optimally Weighted Combination of Multiple Features," Proc. INTERSPEECH 2005, pp.2621-2624, 2005. Florian Eyben, Martin Wollmer, and Bjorn Schuller, "openEAR - Introducing the Munich Open-Source Emotion and Affect Recognition Toolkit," 2009 3rd International Conference on Affective Computing and Intelligent Interaction and Workshops Pedregosa et al., "Scikit-learn: Machine Learning in Python," JMLR 12, pp. 2825-2830, 2011. J.P.Yamron, I.Carp, L.Gillick, S.Lowe, and P.van Mulbregt, “A HIDDEN MARKOV MODEL APPROACH TO TEXT SEGMENTATION AND EVENT TRACKING,” IEEE International Conference on Acoustics, Speech and Signal Processing, pp.333- 336, 1998.

However, since the technique described in the above document uses the behavior history information of the customer, it is not possible to predict the cancellation risk even for a customer who does not have the behavior history information and contacts the customer for the first time.

An object of the present invention is to provide a technique for solving the above problems.

In order to achieve the above object, an information processing device according to the present invention includes:
a means for acquiring a speech amount feature amount relating to the amount of speech of the customer and the operator, calculated based on voice data during a call between the customer and the operator;
means for obtaining an emotion value feature quantity relating to the emotion values of the customer and the operator during the call, calculated based on the voice data;
using a model for predicting the customer's cancellation risk based on the speech amount feature amount and the emotion value feature amount, the acquired speech amount feature amount, and the acquired emotion value feature amount, the customer a means of predicting the churn risk of
Prepare.

In order to achieve the above object, an information processing method according to the present invention comprises:
a step of obtaining a speech amount feature amount relating to the amount of speech of the customer and the operator, calculated based on voice data during a call between the customer and the operator;
a step of obtaining an emotion value feature amount relating to the emotion values of the customer and the operator during the call, calculated based on the voice data;
using a model for predicting the customer's cancellation risk based on the speech amount feature amount and the emotion value feature amount, the acquired speech amount feature amount, and the acquired emotion value feature amount, the customer predicting the churn risk of
including.

In order to achieve the above object, an information processing program according to the present invention comprises:
a step of obtaining a speech amount feature amount relating to the amount of speech of the customer and the operator, calculated based on voice data during a call between the customer and the operator;
a step of obtaining an emotion value feature amount relating to the emotion values of the customer and the operator during the call, calculated based on the voice data;
using a model for predicting the customer's cancellation risk based on the speech amount feature amount and the emotion value feature amount, the acquired speech amount feature amount, and the acquired emotion value feature amount, the customer predicting the churn risk of
run on the computer.

According to the present invention, it is possible to predict the cancellation risk even for a customer who does not have action history information and contacts for the first time.

1 is a block diagram showing the configuration of an information processing device according to a first embodiment of the present invention; FIG. FIG. 3 is a block diagram showing the configuration of an information processing apparatus according to a second embodiment of the present invention; FIG. FIG. 9 is a block diagram showing the configuration of an emotion feature calculation unit of the information processing device according to the second embodiment of the present invention; It is a figure which shows the hardware constitutions of the information processing apparatus which concerns on 2nd Embodiment of this invention. 9 is a flowchart for explaining the processing procedure of the information processing apparatus according to the second embodiment of the present invention; FIG. 11 is a block diagram showing the configuration of an information processing apparatus according to a third embodiment of the present invention; FIG. It is a figure which shows the hardware constitutions of the information processing apparatus which concerns on 3rd Embodiment of this invention. It is a flow chart explaining a processing procedure of an information processor concerning a 3rd embodiment of the present invention.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments for carrying out the present invention will be exemplarily described in detail with reference to the drawings. However, the configuration, numerical values, flow of processing, functional elements, etc. described in the following embodiments are only examples, and modifications and changes are free, and the technical scope of the present invention is not limited to the following description. It is not intended to be limited.

[First Embodiment]
An information processing apparatus 100 as a first embodiment of the present invention will be described using FIG. The information processing device 100 is a device that predicts cancellation risk in a call based on the speech volume feature and emotion in the call.

As shown in FIG. 1 , the information processing device 100 includes a call storage unit 101 , a speech amount feature calculation unit 102 , an emotion recognition unit 103 and a cancellation prediction unit 104 . The call storage unit 101 acquires and stores calls between customers and operators. The speech amount feature calculating unit 102 calculates the speech amount feature of the stored call. The emotion recognition unit 103 recognizes the emotion of the utterance period. The cancellation prediction unit 104 predicts the cancellation risk of the call based on the speech amount feature amount and the recognized emotion.

According to this embodiment, it is possible to predict the cancellation risk even for a customer who does not have action history information and has contacted the customer for the first time.

[Second embodiment]
Next, an information processing apparatus according to a second embodiment of the present invention will be described with reference to FIGS. 2 to 5. FIG. FIG. 2 is a block diagram showing the configuration of the information processing apparatus according to this embodiment.

The information processing apparatus according to the present embodiment can extract customers with high cancellation risk and customers who are expected to continue from calls in call centers and the like, and selectively perform appropriate follow-ups for each customer. It is a device that can be implemented.

The information processing device 200 has a call storage unit 201 , a speech amount feature calculation unit 202 , an emotion recognition unit 203 , an emotion feature calculation unit 204 , a customer extraction model storage unit 205 , a cancellation prediction unit 206 and an extracted customer output unit 207 .

A call storage unit 201 acquires a call between a customer and an operator as call voice data. Then, the call storage unit 201 stores the acquired call voice data as a file or stream data.

When storing the call voice data, the call storage unit 201 assigns a call identifier (call ID) for specifying the call to the call voice data to be stored. Further, the call storage unit 201 detects the speech period of the call sound data from the stored call sound data, and adds speaker information, that is, operator information or customer information to each of the detected speech periods. and memorize it. Furthermore, the call storage unit 201 adds speech time information, that is, the time of the start of speech and the time of the end of the speech to each of the detected speech segments, and stores the speech data to be stored. Therefore, when storing the call voice data, the call storage unit 201 adds a call ID, operator information, customer information, and utterance time information to the call voice data to be stored.

Note that various methods have been disclosed for detecting speech segments, and for example, the method described in Non-Patent Document 1 may be used.

As for the call ID, for example, in recent call centers, the call ID can be acquired by a CTI (Computer Telephony Integration) system. It is also possible to record data on separate channels for the customer side signal and the operator side signal. Therefore, the call ID and speaker information can be assigned using the call ID and channel information obtained by the CTI system.

Also, the speech time information can be given by acquiring the absolute time of the detected speech section or the time information defined by the system for the beginning and end.

The speech amount feature calculation unit 202 calculates the feature of the amount of speech between the operator and the customer from the time information of each speech period stored in the call storage unit 201 . That is, the total time of speech segments of the operator and the customer in each call, the ratio of the total time of speech segments of the operator and the customer in each call, the average length of the speech segments of the operator and the customer in each call, and the variance At least one of a value, a median, a mode, or a value calculated by other statistical processing is used.

Emotion recognition unit 203, for each utterance segment stored in call storage unit 201, valence (emotional valence indicating the quality of emotion, degree of positive or negative emotion), arousal (arousal level indicating arousal or calmness, Emotional excitement level), at least one emotional value of satisfaction and expectation is calculated. Various emotion recognition methods have been disclosed, and for example, the method described in Non-Patent Document 2 can be used.

Each emotion is calculated using a predetermined range of numeric values or a finite number of categories as an emotion value. That is, the emotion recognition unit 203 calculates, for example, a value in the range of -3 to 3 or a category such as {-3, -2, -1, 0, 1, 2, 3} for valence. The same is true for arousal, conviction, and expectation, but the ranges and categories specified for each emotion do not have to be the same.

The emotion feature calculation unit 204 calculates the emotion feature amount of the operator and the customer from the emotion value of each utterance period calculated by the emotion recognition unit 203 . Emotion features include the sum of the emotion values of the operator and the customer in each call, the emotion value ratio, the average value of the emotion values, the variance value, the median value, the mode value, or other statistical processing. At least one of the values is included.

FIG. 3 is a block diagram showing the configuration of the emotion feature calculator of the information processing apparatus according to this embodiment. The emotion feature calculation unit 204 may calculate the emotion feature amounts of the operator and the customer using only the emotion value of each utterance section included in a specific section in the call voice data, using the configuration shown in FIG. The emotion feature calculator 204 shown in FIG. 3 includes a topic section detector 241 , a narrowing section 242 , and a feature calculator 243 .

The topic segment detection unit 241 detects topic segments included in the call voice data stored in the call storage unit 201, and assigns topic start time and topic end time to each of the detected topic segments. Topics include, for example, openings, closings, product/service explanations, fee explanations, solicitations, interviews, confirmation of personal information, administrative procedures, and the like.

Various methods have been disclosed for detecting topic segments, and for example, speech recognition and the method described in Non-Patent Document 4 can be used in combination. Specifically, it is possible to divide a word string obtained by speech recognition into a plurality of topic sections using a topic model in which word appearance frequencies for various topics are learned in advance.

Also, the topic section detection unit 241 may detect topic sections based on time information or the number of utterances. For example, the first 30-second section or the first 5 utterance sections of a call can be detected as the opening topic section. In addition, a 30-second segment at the end of a call or a 5-speech segment at the end of a call can be detected as a closing topic segment. Furthermore, the latter ⅓ section of the call (for example, the section of 6 to 9 minutes in the case of a 9-minute call) may be detected as one topic section.

Narrowing down section 242 outputs only the speech sections included in the specific topic detected by topic section detection section 208 to feature calculation section 243 among the speech sections to which emotion recognition section 203 assigns an emotion value. For example, only the utterance segments included in the closing topic segment and their emotion values are output to the feature calculation unit 243 .

The feature calculation unit 243 calculates the emotion feature amount of the operator and the customer from the emotion value of each utterance period output from the narrowing-down unit 242 . Emotion features include the sum of emotion values of operators and customers in a specific topic section, the ratio of emotion values, the average value of emotion values, the variance value, the median value, the mode value, or other statistical processing. contains at least one of the values For example, the valence of the customer's utterance and the sum or average value of emotion values of expectations included in the opening or closing topic section are calculated as emotion feature amounts.

The customer extraction model storage unit 205 uses the emotion feature amount of the operator and the customer calculated by the emotion feature calculation unit 204 and the speech amount of the operator and the customer calculated by the speech amount feature calculation unit 202 as feature amounts. A model predicted by a high-risk customer is stored as a customer extraction model.

For the customer extraction model, correct label information indicating whether or not the call has been canceled corresponding to each call data is given in advance, and a model for predicting the cancellation risk is learned by a machine learning method. Various machine learning techniques have been disclosed, and for example, Non-Patent Document 3 discloses details and an executable program thereof. For churn prediction, a prediction method corresponding to the learning method of the model can be used.

The cancellation prediction unit 206 predicts (calculates) the cancellation risk for each call using a customer extraction model.

The extracted customer output unit 207 compares the predicted churn risk of each call with a predetermined churn risk threshold value and a high churn risk rate among the number of target calls, and identifies customers with a high churn risk, i.e., customers with a high churn risk. Extract customers with high calls.

In addition, in the description of the present embodiment, an example of predicting cancellation has been described, but the present embodiment can be similarly applied to prediction of continuation.

FIG. 4 is a block diagram illustrating the hardware configuration of the information processing device 200 according to this embodiment. A CPU (Central Processing Unit) 410 is a processor for arithmetic control, and implements the functional components of the information processing apparatus 200 in FIG. 2 by executing a program. The CPU 410 may have multiple processors and execute different programs, modules, tasks, threads, etc. in parallel. A ROM (Read Only Memory) 420 stores fixed data such as initial data and programs, and other programs. Also, the network interface 430 communicates with other devices and the like via a network. Note that the number of CPUs 410 is not limited to one, and may include a plurality of CPUs or a GPU (Graphics Processing Unit) for image processing. Moreover, it is desirable that the network interface 430 has a CPU independent of the CPU 410 and writes or reads transmission/reception data in a RAM (Random Access Memory) 440 area. It is also desirable to provide a DMAC (Direct Memory Access Controller) for transferring data between RAM 440 and storage 450 (not shown). Further, CPU 410 recognizes that data has been received or transferred to RAM 440 and processes the data. Also, the CPU 410 prepares the processing result in the RAM 440, and entrusts subsequent transmission or transfer to the network interface 430 or DMAC.

RAM 440 is a random access memory used by CPU 410 as a work area for temporary storage. The RAM 440 has an area for storing data necessary for implementing the present embodiment. A call ID 441 is data identifying a call between an operator and a customer, and is assigned to each call. The speech volume feature 442 is the speech volume feature calculated in the call between the operator and the customer. The emotion value 443 is the emotion value calculated for each utterance segment in each call. The emotional feature 444 is a feature of emotion calculated from the emotional value of each utterance interval. The customer extraction model 445 is a model for predicting customers with a high cancellation risk. Churn Prediction 446 is a predicted churn risk prediction for each call using a customer extraction model.

The input/output data 447 is data input/output via the input/output interface 460 . Transmitted/received data 448 is data transmitted/received via network interface 430 . The RAM 440 also has an application execution area 449 for executing various application modules.

The storage 450 stores a database, various parameters, or the following data or programs necessary for realizing this embodiment. The storage 450 stores a speech amount feature calculation module 453 , an emotion recognition module 454 , an emotion feature calculation module 455 and a cancellation prediction module 457 .

The speech amount feature calculation module 453 is a module for calculating the speech amount feature of a call between an operator and a customer. The emotion recognition module 454 is a module that calculates the emotion value in the utterance section of the call between the operator and the customer. The emotion feature calculation module 455 is a module that calculates the feature of emotion from the calculated emotion value of the utterance period. The cancellation prediction module 457 is a module that predicts the cancellation risk of a call based on the feature amount of speech and the recognized emotion. These modules 453 to 457 are read by the CPU 410 into the application execution area 449 of the RAM 440 and executed. The control program 456 is a program for controlling the information processing apparatus 200 as a whole.

The input/output interface 460 interfaces input/output data with input/output devices. A display unit 461 and an operation unit 462 are connected to the input/output interface 460 . A storage medium 464 may also be connected to the input/output interface 460 . Furthermore, a speaker 463 as an audio output unit, a microphone (not shown) as an audio input unit, or a GPS position determination unit may be connected. Note that the RAM 440 and the storage 450 shown in FIG. 4 do not show programs or data relating to general-purpose functions of the information processing apparatus 200 or other realizable functions.

FIG. 5 is a flowchart for explaining the processing procedure of the information processing apparatus 200 according to this embodiment. This flowchart is executed by the CPU 410 in FIG. 4 using the RAM 440, and implements the functional components of the information processing apparatus 200 in FIG.

In step S501, the information processing device 200 acquires a call between an operator and a customer. In step S503, the information processing apparatus 200 detects speech periods of the acquired call, and provides speaker information and speech time information to each detected speech period. In step S505, the information processing apparatus 200 calculates the feature of the amount of speech between the operator and the customer in the call. In step S507, the information processing apparatus 200 calculates the emotion value of the utterance period. In step S509, the information processing apparatus 200 calculates an emotion feature from the calculated emotion value of the utterance period. In step S513, the information processing device 200 predicts the cancellation risk for each call using the customer extraction model. In step S515, the information processing apparatus 200 extracts call customers with a high cancellation risk based on the predicted cancellation risk.

According to this embodiment, it is possible to predict cancellation even for customers whose behavior history cannot be obtained, such as contacting customers for the first time, by using the customer's inquiries about products and services and call data at the time of application. .

In addition, according to the present embodiment, information such as which of the operator or the customer was actively conversing is calculated by using statistical processing, including information that cannot be ruled manually, as an utterance feature. Similarly, a model that estimates churn risk with high accuracy by calculating using statistical processing including information that cannot be ruled manually from emotional values, rather than emotional values that can be interpreted by humans such as satisfaction and dissatisfaction as emotional characteristics. can be learned and predicted.

Whether to cancel or not to cancel is not decided by the customer's inquiry about products and services or the interaction at the time of application. Some customers tend to cancel when they purchase products that they did not originally intend through upselling or cross-selling. In such a case, the present inventor believes that the conversational information of inquiries to services and exchanges at the time of application are effective in predicting cancellation or continuation, and that utterance features and emotional features are effective at that time. This is what I discovered.

Therefore, in the present embodiment, the predicted (calculated) risk is compared with a pre-designated cancellation risk threshold and the high cancellation risk rate in the number of target calls, and customers judged to have a high cancellation risk are extracted. . As a result, in the present embodiment, it is possible to extract customers who have a high risk of canceling a contract mainly due to telephone correspondence.

Also, the inventors have found that whether or not a customer will cancel in the future depends on the operator's or customer's emotions during a particular topical segment in a call. For example, we found that the customer's positive emotions and expectations in the topical section of the opening and closing are high, the cancellation risk is low.

Therefore, in the present embodiment, the cancellation risk is estimated based on the emotion feature focusing on the emotion of the operator or the customer in a specific topic section of the call. Thereby, in the present embodiment, it is possible to extract customers with a high cancellation risk with high accuracy.

[Third embodiment]
Next, an information processing apparatus according to a third embodiment of the present invention will be explained using FIGS. 6 to 9. FIG. FIG. 6 is a block diagram showing the configuration of the information processing apparatus according to this embodiment. The information processing apparatus according to this embodiment differs from that of the second embodiment in that it has a phrase storage unit, a phrase recognition unit, and a phrase feature calculation unit. Since other configurations and operations are similar to those of the second embodiment, the same configurations and operations are denoted by the same reference numerals, and detailed description thereof will be omitted.

The information processing device 600 has a phrase storage unit 601, a phrase recognition unit 602, and a phrase feature calculation unit.

The phrase storage unit 601 stores predesignated phrases. The phrase storage unit 601 stores phrases that can be regarded as situations of at least one of satisfaction, dissatisfaction, brute force, and attention to fees. For example, in the case of satisfaction, it is "understood", "that's right", and the like. In the case of dissatisfaction, it is "for the time being", "for the time being", "for the time being", and the like. In the case of brute force, it is "Let's decide", "This is fine", and so on. In the case of sticking to the price, it is "price", "fee", "free shipping", "consumption tax", "mottainai", and the like.

A phrase recognition unit 602 detects and recognizes the appearance of phrases stored in the phrase storage unit 601 . A phrase recognition unit 602 performs speech recognition and word spotting, detects the appearance of phrases stored in the phrase storage unit 601, and recognizes them.

The phrase feature calculator 603 calculates features of phrases of operators and customers. The phrase feature calculation unit 603 uses the recognition result (detection result) by the phrase recognition unit 602 to perform the following calculations for at least one of satisfaction, dissatisfaction, aggressiveness, and commitment to fees. In other words, the phrase feature calculation unit 603 calculates the number of occurrences of phrases stored in the phrase storage unit 601 . The phrase feature calculation unit 603 also calculates the normalized number of phrase appearances obtained by normalizing the number of phrase appearances by the utterance time, and the number of phrase appearances for each predetermined time interval. The phrase feature calculation unit 603 calculates at least one of the number of phrase appearances, the normalized number of phrase appearances, and the number of phrase appearances for each predetermined time interval.

As in the second embodiment, the customer extraction model storage unit 205 stores the emotion feature amounts of the operator and the customer calculated by the emotion feature calculation unit 204 and the utterances of the operator and the customer calculated by the utterance amount feature calculation unit 202. A model is generated by adding the following feature quantity to the quantity feature. That is, the customer extraction model storage unit 205 adds the phrase feature amounts of the operator and the customer calculated by the phrase feature calculation unit 603 to the emotion feature amount and the utterance amount feature to predict customers with a high cancellation risk. Generate and store an extraction model. The learning (generation) of the customer extraction model and the prediction using the learned model are the same as in the second embodiment except that the above-described feature amount (phrase feature amount) is used.

FIG. 7 is a block diagram illustrating the hardware configuration of the information processing device 600 according to this embodiment. A CPU (Central Processing Unit) 410 is a processor for arithmetic control, and implements the functional components of the information processing apparatus 600 in FIG. 2 by executing a program.

RAM 740 is a random access memory used by CPU 410 as a work area for temporary storage. The RAM 740 has an area for storing data necessary for implementing the present embodiment. Stored phrases 741 are predesignated and stored phrases.

The storage 750 stores a database, various parameters, or the following data or programs necessary for implementing this embodiment. Storage 850 stores phrase 751 , phrase recognition module 752 and phrase feature calculation module 753 . Phrase 751 is a predesignated phrase. The phrase recognition module 752 is a module that detects and recognizes the appearance of stored phrases by voice recognition or word spotting. The phrase feature calculation module 753 uses the phrase recognition result (detection result) to calculate at least one of the number of occurrences of phrases, the number of occurrences of phrases normalized by the utterance time, and the number of occurrences of phrases at predetermined time intervals. It is a module that calculates. These modules 752 to 753 are read by the CPU 410 into the application execution area 449 of the RAM 840 and executed.

FIG. 8 is a flowchart for explaining the processing procedure of the information processing apparatus 600 according to this embodiment. This flowchart is executed by the CPU 410 in FIG. 7 using the RAM 840, and implements the functional components of the information processing apparatus 600 in FIG.

At step 801, the information processing device 600 detects and recognizes the appearance of a predesignated phrase. In step S803, the information processing apparatus 600 calculates features of the recognized phrase. The features of phrases are, for example, the number of occurrences of phrases, the number of occurrences of phrases normalized by the utterance time, the number of occurrences of phrases for each predetermined time interval, and the like.

According to the present embodiment, relevant phrases are detected for consent, dissatisfaction, aggressiveness, and commitment to fees, and phrase features are calculated, and a model is generated by adding these phrase features. can be predicted with high accuracy by taking into account the content of the utterance.
[Other embodiments]
Although the present invention has been described with reference to the embodiments, the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. Also, any system or apparatus that combines separate features included in each embodiment is also included in the scope of the present invention.

Further, the present invention may be applied to a system composed of a plurality of devices, or may be applied to a single device. Furthermore, the present invention is also applicable when an information processing program that implements the functions of the embodiments is directly or remotely supplied to a system or apparatus. Therefore, in order to implement the functions of the present invention on a computer, a program installed in a computer, a medium storing the program, and a WWW (World Wide Web) server from which the program is downloaded are also included in the scope of the present invention. . In particular, non-transitory computer readable media containing programs that cause a computer to perform at least the processing steps included in the above-described embodiments are included within the scope of the present invention.

[Other expressions of the embodiment]
Some or all of the above embodiments can also be described as the following additional remarks, but are not limited to the following.
(Appendix 1)
call storage means for capturing and storing calls between the customer and the operator;
speech amount feature calculation means for calculating the stored speech amount feature amount of the call;
emotion recognition means for recognizing the emotion of the utterance section;
cancellation prediction means for predicting a cancellation risk in the call based on the utterance amount feature amount and the recognized emotion;
Information processing device with
(Appendix 2)
The emotion recognition means further calculates an emotion value of the recognized emotion,
emotion feature calculation means for calculating an emotion feature, which is a feature of an emotion, from the emotion value;
customer extraction model storage means for storing pre-learned customer extraction models for predicting and extracting customers with a high cancellation risk based on the utterance amount feature amount and the emotion feature;
further comprising
The information processing apparatus according to appendix 1, wherein the cancellation prediction means predicts the cancellation risk based on the customer extraction model.
(Appendix 3)
The emotion feature calculation means is
topic section detection means for detecting a specific topic section included in a call;
narrowing means for outputting the emotion value of the utterance segment included in the specific topic segment;
feature calculation means for calculating an emotion feature, which is a feature of an emotion, from the emotion value output by the narrowing-down means;
The information processing apparatus according to appendix 2, further comprising:
(Appendix 4)
a phrase storage means for storing a predesignated phrase;
Phrase recognition means for detecting and recognizing the occurrence of phrases stored in the phrase storage means in the call;
Phrase feature calculation means for calculating phrase features that are features of phrases that have appeared;
further comprising
4. The information processing apparatus according to any one of appendices 1 to 3, wherein the cancellation prediction means further predicts the cancellation risk based on the phrase features.
(Appendix 5)
5. Any one of Appendices 1 to 4, wherein the call storage means detects an utterance period of the call, adds speaker information and utterance time information to the detected utterance period, and stores the utterance. The information processing device according to item 1.
(Appendix 6)
6. The information processing apparatus according to any one of appendices 1 to 5, wherein the emotion includes at least one of valence, arousal, satisfaction, and expectation.
(Appendix 7)
The topic section includes at least one of opening, closing, product/service explanation, fee explanation, solicitation, hearing, personal information confirmation and administrative procedure in the call,
6. The information processing apparatus according to any one of appendices 3 to 5, wherein the emotion includes at least one of valence and expectation.
(Appendix 8)
8. The information processing apparatus according to any one of Appendices 2 to 7, wherein the emotion feature includes at least one of an emotion value ratio of the emotion values, an average value, a variance value, a median value, and a mode.
(Appendix 9)
The phrase features include at least one of the number of phrase appearances, which is the number of appearances of the phrase, the normalized number of phrase appearances obtained by normalizing the number of phrase appearances by the utterance time, and the number of phrase appearances for each predetermined time interval. 8. The information processing apparatus according to any one of items 1 to 7.
(Appendix 10)
a call storage step of retrieving and storing calls between the customer and the operator;
an utterance amount feature calculation step of calculating the stored utterance amount feature amount of the call;
an emotion recognition step of recognizing the emotion of the utterance segment;
a cancellation prediction step of predicting a cancellation risk in the call based on the utterance amount feature quantity and the recognized emotion;
Information processing method including.
(Appendix 11)
a call storage step of retrieving and storing calls between the customer and the operator;
an utterance amount feature calculation step of calculating the stored utterance amount feature amount of the call;
an emotion recognition step of recognizing the emotion of the utterance segment;
a cancellation prediction step of predicting a cancellation risk in the call based on the utterance amount feature quantity and the recognized emotion;
An information processing program that causes a computer to execute

Claims (10)

a means for acquiring a speech amount feature amount relating to the amount of speech of the customer and the operator, calculated based on voice data during a call between the customer and the operator;
means for obtaining an emotion value feature quantity relating to the emotion values of the customer and the operator during the call, calculated based on the voice data;
using a model for predicting the customer's cancellation risk based on the speech amount feature amount and the emotion value feature amount, the acquired speech amount feature amount, and the acquired emotion value feature amount, the customer a means of predicting the churn risk of
Information processing device. The speech data includes speaker information indicating whether the speaker is the customer or the operator , the start time of the speech, and the time of the end of the speech, for the speech data of the detected speech period. Time information and are given,
the means for acquiring the utterance amount feature acquires the utterance amount feature calculated using the speaker information and the utterance time information;
2. The information processing apparatus according to claim 1 , wherein said means for acquiring said emotion value feature quantity acquires said emotion value feature quantity calculated using the emotion value in each of said utterance intervals. The utterance amount feature is
the total time of speech of the customer and the operator; and
a ratio of the total time of the customer's speech to the total time of the operator's speech;
an average value of the length of the customer's speech interval and an average value of the length of the speech interval of the operator;
A variance value of the length of the speech period of the customer and a variance value of the length of the speech period of the operator;
a median length of the utterance interval of the customer and a median length of the utterance interval of the operator;
a mode of the length of the utterance segment of the customer and a mode of the length of the utterance segment of the operator;
3. The information processing apparatus according to claim 1, comprising at least one of: 4. The emotional value according to any one of claims 1 to 3, wherein the emotional value includes at least one of valence (emotional valence indicating the quality of emotion) , arousal (arousal indicating arousal or calmness) , satisfaction, and expectation. The information processing device described. 5. The information processing apparatus according to any one of claims 1 to 4, wherein the emotion value feature amount includes at least one of a ratio of the emotion values, an average value, a variance value, a median value, and a mode value. 4. The information processing apparatus according to claim 2, wherein said means for acquiring said emotion value feature quantity acquires said emotion value feature quantity calculated based on voice data of a specific topic section of said call. The specific topic section includes the opening topic section , the closing topic section , the product/service explanation topic section , the charge explanation topic section , the solicitation topic section , the hearing topic section , and the personal information confirmation section. 7. The information processing apparatus according to claim 6 , comprising at least one of a topic section and a topic section of paperwork. Further comprising means for acquiring phrase features of the customer and the operator during the call, which are calculated based on the voice data ;
The means for predicting includes a model for predicting the customer's cancellation risk based on the speech amount feature amount, the emotion value feature amount, and the phrase feature amount, the acquired speech amount feature amount , and the acquired emotion Predicting the customer's cancellation risk using the value feature amount and the acquired phrase feature amount ,
8. The information processing apparatus according to any one of claims 1 to 7 , wherein said phrase feature amount is calculated by recognizing occurrence of a pre-specified phrase during said call . a step of obtaining a speech amount feature amount relating to the amount of speech of the customer and the operator, calculated based on voice data during a call between the customer and the operator;
a step of obtaining an emotion value feature amount relating to the emotion values of the customer and the operator during the call, calculated based on the voice data;
using a model for predicting the customer's cancellation risk based on the speech amount feature amount and the emotion value feature amount, the acquired speech amount feature amount, and the acquired emotion value feature amount, the customer predicting the churn risk of
Information processing method including. a step of obtaining a speech amount feature amount relating to the amount of speech of the customer and the operator, calculated based on voice data during a call between the customer and the operator;
a step of obtaining an emotion value feature amount relating to the emotion values of the customer and the operator during the call, calculated based on the voice data;
using a model for predicting the customer's cancellation risk based on the speech amount feature amount and the emotion value feature amount, the acquired speech amount feature amount, and the acquired emotion value feature amount, the customer predicting the churn risk of
An information processing program that causes a computer to execute

Images