JP2013211926A - Abnormal conversation detecting apparatus, abnormal conversation detecting method, and abnormal conversation detecting program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an abnormal conversation detecting apparatus and the like which accurately detect abnormal conversation without being influenced by individual characteristics of an operator or an intended change in conversation.SOLUTION: The abnormal conversation detecting apparatus includes: a conversation time measuring section 310 for measuring the speaking time of an operator 102 and the speaking time of the calling opposite party of the operator 102; a conversation information analyzing section 320 for calculating a ratio of speaking by the operator 102 or that by the opposite party on the basis of each speaking time measured by the conversation time measuring section 310; and means for outputting on a display 109a display corresponding to a result of comparing the ratio calculated by the speaking information analyzing section 320 with the ratio of the speaking to be compared.

Description

  The present invention relates to a conversation abnormality detection device that detects a conversation abnormality between an operator and a communication partner, and more particularly to a conversation abnormality detection device that detects a conversation abnormality based on a conversation time.

  In the transmission type and reception type call centers, a system in which the manager monitors conversation exchanges between the “telephone operator” and the “destination” is often used for the purpose of monitoring the occurrence of complaints and instructing telephone correspondence. Specifically, when a complaint occurs, an administrator who has been monitoring goes to the operator's seat and gently gives a note of the response method to the operator. Also, a general method is to temporarily hold a conversation or close a telephone opening with a hand and resume the conversation after consulting with a manager about a handling method.

  This method has the following problems. When multiple operators are talking at the same time, it is impossible to monitor all of them. For this reason, it largely depends on the manager's experience and intuition (operators, new operators, etc., who frequently generate complaints), and it becomes difficult to find signs of complaints.

  Regarding such a problem, a technique for determining whether or not the operator is hesitant based on the voice information of the operator and assisting the operator when he / she is hesitant is disclosed (see Patent Document 1). ).

JP 2007-4000

  However, the technique disclosed in Patent Document 1 is based on voice information (basic frequency, sound pressure level) obtained from a talking operator, feature information (interconnection time, ratio between them), etc. Is evaluating the degree of confusion. Therefore, when the operator changes his / her personal characteristics (for example, uprising, stuttering, inflection, etc.), intentional voice changes of the operator, changes in the way of speaking, and during conversations with other operators Etc., and there is a problem that accurate information cannot be obtained.

  Accordingly, the present invention has been made to solve the above-described problems, and is a conversation abnormality detection device that accurately detects a conversation abnormality without being influenced by the personal characteristics of the operator or intentional changes in conversation. An object of the present invention is to provide a conversation abnormality detection method and a conversation abnormality detection program.

  The conversation abnormality detection device disclosed in the present application is based on conversation time measuring means for measuring the utterance time of the operator and the utterance time of the operator's call partner, and each utterance time measured by the conversation time measurement means. Conversation information analyzing means for calculating the proportion of utterances by the operator or the proportion of utterances by the other party, and display according to the result of comparing the proportion calculated by the conversation information analyzing means and the proportion of utterances to be compared. Output means.

  Although the present invention has been shown as an apparatus so far, as will be apparent to those skilled in the art, the present invention can also be understood as a method and a program. These outlines of the invention do not enumerate the features essential to the present invention, and a sub-combination of these features can also be an invention.

  In other words, a method, a circuit, a system, a computer program, a recording medium, a data structure, etc., in which a component or a combination of components of the conversation abnormality detection device disclosed in the present application is applied is also effective as another aspect. .

It is a system outline figure of an operator management system using a conversation abnormality detecting device concerning a 1st embodiment. It is a hardware block diagram of the conversation abnormality detection apparatus which concerns on 1st Embodiment. It is a functional block diagram of the conversation abnormality detection apparatus which concerns on 1st Embodiment. It is a flowchart which shows operation | movement of the conversation abnormality detection apparatus which concerns on 1st Embodiment. It is a figure which shows an example of the reference | standard information of a conversation ratio. It is a figure which shows the measurement result of conversation time. It is a figure which shows an example of the notification screen to an administrator when it determines with a conversation being abnormal. It is a functional block diagram of the conversation abnormality detection apparatus which concerns on 2nd Embodiment. It is a flowchart which shows operation | movement of the conversation abnormality detection apparatus which concerns on 2nd Embodiment. It is a figure which shows the process in the case of correct | amending conversation time.

  Embodiments of the present invention will be described below. The present invention can be implemented in many different forms. Therefore, the present invention should not be construed based only on the description of the present embodiment. Also, the same reference numerals are given to the same elements throughout the present embodiment.

  In the following embodiments, the apparatus will be mainly described. However, as is apparent to those skilled in the art, the present invention can also be implemented as a method and a program for operating a computer. In addition, the present invention can be implemented in hardware, software, or hardware and software embodiments. The program can be recorded on any computer-readable medium such as a hard disk, CD-ROM, DVD-ROM, optical storage device, or magnetic storage device. Furthermore, the program can be recorded on another computer via a network.

(First embodiment of the present invention)
The conversation abnormality detection device according to the present embodiment will be described with reference to FIGS.
(1. Configuration)
Below, the structure of the conversation abnormality detection apparatus which concerns on this embodiment is demonstrated.
FIG. 1 is a system outline diagram of an operator management system using the conversation abnormality detection device according to the present embodiment. The operator 102 has a call function and has a conversation with the customer 101 by telephone using a device that stores and displays various information related to the conversation. The conversation abnormality detection device 300 detects whether or not there is an abnormality while recording the conversation between the operator 102 and the customer 101. When an abnormality is detected, the fact that an abnormality has occurred is displayed on the monitoring monitor used by the administrator 103. The administrator 103 looks at the display indicating that an abnormality has occurred and gives appropriate advice to an operator who seems to have an abnormality.

FIG. 2 is a hardware configuration diagram of the conversation abnormality detection device according to the present embodiment. The computer 1 includes a CPU (Central Processing Unit) 101, a RAM (Random Access Memory) 102, a ROM (Read Only Memory) 103, a flash memory (Flash Memory) 104, and an HD (Hard Disk) 105 that is an external storage device. A LAN (Local Area Network) card 106, a mouse 107, and a keyboard 108 are also provided. Furthermore, a video card 109, a display 109a which is a display device electrically connected to the video card 109, a sound card 110, and a speaker 110a which is a sound output device electrically connected to the sound card 110 are provided. Furthermore, a drive 111 for reading and writing storage media such as a flexible disk, a CD-ROM, and a DVD-ROM is provided.
Note that the above hardware configuration is merely an example, and it is natural that the components can be changed.

  FIG. 3 is a functional block diagram of the conversation abnormality detection device according to the present embodiment. The conversation abnormality detection device 300 includes a conversation time measurement unit 310, a conversation information analysis unit 320, a conversation abnormality determination unit 330, and a reference information unit 340. The conversation information analysis unit 320 includes a conversation rate calculation unit 321 and an average value calculation unit 322.

  The conversation time measuring unit 310 performs processing for measuring each conversation time and the like for conversation information in the conversation between the operator 102 and the customer 101 input as the input data 301. When there are a plurality of operators 102, the conversation time and the like are measured for all operators 102 and customers 101.

  The conversation information analysis unit 320 performs processing for analyzing conversation information based on the conversation time measured by the conversation time measurement unit 310. In addition to the ratio between the conversation time of the operator 102 and the conversation time of the customer 101 and the average value of the conversation information shown below, the length of the conversation time, the number of questions and answers within a certain time, The time during which the voice is uttered and the time during which there is no voice may be analyzed.

  The conversation ratio calculation unit 321 performs processing for calculating the ratio between the conversation time of the operator 102 and the conversation time of the customer 101 based on the conversation time measured by the conversation time measurement unit 310.

  The average value calculation unit 322 is based on the conversation time measured by the conversation time measurement unit 310, the average conversation time of the plurality of operators 102, the average conversation time of the customer 101, and the conversation between the operator 102 and the customer 101. A process for calculating an average value of the ratio of time and the like is performed.

  The conversation abnormality determination unit 330 compares the information analyzed by the conversation information analysis unit 320 with the reference information stored in the reference information unit 340 to determine whether there is an abnormality in the conversation, and the determination result Is displayed on the display 109a. When there are a plurality of operators 102, the average values such as conversation time and conversation ratio calculated by the average value calculation unit 322 are compared with the conversation time and conversation ratio of each operator 102. Then, regarding the conversation information of the operator 102 that is largely deviated from the average value, it is determined that an abnormality has occurred in the conversation of the operator 102.

The reference information unit 340 is a data unit that stores the reference information related to the conversation time input as the input data 301 for each job. Reference information includes information such as the length of the conversation time, the ratio between the conversation time of the operator 102 and the conversation time of the customer 101, the number of questions and answers within a certain period of time, the time of utterance at the same time, the time of no utterance simultaneously Is included.
Examples of each business include a dunning business, a sales promotion business, a complaint processing business, an order reception business, an inquiry response business, and the like.

(2. Operation)
Below, operation | movement of the conversation abnormality detection apparatus which concerns on this embodiment is demonstrated.
FIG. 4 is a flowchart showing the operation of the conversation abnormality detection device according to the present embodiment. First, reference information is input as input data 301 and stored in the reference information unit 340 (step S401).
The reference information may be directly input from the mouse 207, the keyboard 208, or the like, or the reference information stored in an external storage medium may be read from the drive 211 and input.

  Here, the reference information will be described in detail. FIG. 5 is a diagram showing an example of the conversation rate reference information. The upper part of FIG. 5 is reference information indicating a normal value, and the lower part is reference information indicating an abnormal value. Further, the left column is reference information about the dunning business, and the right column is standard information about the sales promotion (telemarketing) business.

  The reference information includes reference ratio information, and the reference ratio information is information that the conversation ratio between the conversation time of the operator 102 and the conversation time of the customer 101 is m: n. This reference ratio information has different characteristics for each business.

  For example, in the reminding work, since the content to be communicated is determined, the story is less likely to be detoured, and it is the main duty to attach the consent or promise of the customer 101. It is common for the amount of conversation to be small. Therefore, the conversation time for one session is 3 to 5 minutes, and the conversation ratio is about 6: 4.

  Further, for example, in the sales promotion business, when negotiations are proceeding well, conversations regarding questions and requests are made with respect to the explanation of the operator 102, so the amount of conversation is the same or the number of customers 101 slightly increases. Therefore, the conversation time is 5 to 10 minutes and the conversation ratio is about 4: 6.

  In the reference information unit 340, a threshold value for setting a normal value range is set and stored based on the reference information as the normal value. For example, as shown in FIG. 5, in the case of a reminder service, a conversation has occurred after 5 minutes, and the conversation is abnormal when the conversation time of the customer 101 is more than twice the conversation time of the operator 102. there is a possibility. Therefore, a conversation time of 5 minutes and a conversation ratio of 1: 2 (m = 1, n = 2) are set as threshold values. When the threshold range is exceeded (when the conversation time is longer than 5 minutes and n> 2 for m = 1), it is determined that the conversation is abnormal.

  Further, for example, as shown in FIG. 5, in the case of a sales promotion business, when the conversation time of the operator 102 becomes twice or more than the conversation time of the customer 101 even though the conversation time is long to some extent, one-sided sales etc. May be an unusual conversation. Therefore, a conversation time of 10 minutes and a conversation ratio of 2: 1 (m = 2, n = 1) are set as threshold values. When this threshold is exceeded (when the conversation time is longer than 10 minutes and m> 2 for n = 1), it is determined that the conversation is abnormal.

Further, when there are a plurality of operators 102, the average value of the conversation information is compared with the conversation information of each operator 102 to determine whether the conversation is abnormal. The threshold value in that case is also set and stored. For example, it is assumed that the reference ratio information as a normal value is set to 6: 4. If there are 100 operators and the average value of the overall conversation ratio is 7: 3, if the conversation ratio is 5: 5 for a certain operator 102, the one It is determined that the conversation of the operator 102 is abnormal.
It is assumed that the threshold range is set to 10% (if the reference ratio information is 7: 3, the normal value is between 6: 4 and 8: 2).

Based on the preset reference information, if the conversation ratio is 5: 5, it is determined as a normal value (because the reference ratio information is set to 6: 4), but the overall average value is used as the reference information. When priority is given, it is determined as an abnormal value (because the average value of the overall conversation rate is 7: 3). Which reference information is given priority can be arbitrarily set by the user, and may be determined using both pieces of reference information.
Note that the reference information shown in FIG. 5 is merely an example, and the reference information and the threshold value may be freely set according to the environment and situation.

Returning to FIG. 4, when the reference information is stored, the conversation time of the operator 102 and the customer 101 is measured (step S402).
Here, the measurement of the conversation time will be described. FIG. 6 is a diagram illustrating a measurement result of the conversation time. One square is 10 seconds, and when there is a conversation, ■, and when there is no conversation, □. The total conversation time is 100 seconds, the conversation of the operator 102 is performed between 1 to 20 seconds, 51 seconds to 70 seconds, and the conversation of the customer 101 is 11 seconds to 60 seconds, 71 seconds to 80 seconds, It shows that it is performed between 91 seconds and 100 seconds. Also, both indicate conversation between 11 seconds and 20 seconds, and neither indicate conversation between 81 seconds and 90 seconds. As a case where neither of them is talking, for example, it can be considered that the state was on hold. Further, when both are talking, both may talk alternately during 10 seconds. This is the case, for example, when one has a conversation for 3 seconds and the other has a conversation for 7 seconds. On the other hand, the operator 102 and the customer 101 may speak at the same time. For example, the case where the operator has a conversation from 11 seconds to 18 seconds and the customer from 15 seconds to 20 seconds can be considered. At this time, between 102 seconds and 18 seconds, the operator 102 and the customer 101 are simultaneously covering and speaking.

  In the present embodiment, the time that the operator 102 utters and the time that the customer 101 utters are accurately measured as the conversation time. In other words, even if the voice is put on at the same time, it is accumulated in each conversation time.

  Note that in order to accurately measure the time when the operator 102 uttered and the time when the customer 101 uttered, the characteristics (frequency, range, sound pressure, conversation, etc.) of each voice are extracted, You may make it distinguish from a holding tone. As described above, by accurately measuring the time when the operator 102 utters and the time when the customer 101 utters, a more accurate conversation time can be obtained and an abnormal conversation can be detected accurately. become.

  At this time, in addition to the overall conversation time, the conversation time of the operator 102, and the conversation time of the customer 101, the time spoken at the same time, the time required for one question and answer in a series of conversations, and the question and answer at a predetermined time You may make it also measure information, such as the frequency | count of having performed.

  Returning to FIG. 4, when the conversation time is measured, the conversation ratio is calculated based on the measured conversation time (step S403). As described above, the conversation ratio is a ratio between the conversation time of the operator 102 and the conversation time of the customer 101. The conversation ratio is calculated, and the average value of the conversation information (conversation time, conversation ratio, etc.) of all operators is calculated (step S404).

  The conversation ratio calculated in step S403 is compared with the reference information stored in the reference information unit 340 (step S405). Further, the average value calculated in step S404 is compared with the conversation information of each operator 102 (step S406). As a result of the comparison in step S405 and step S406, it is determined whether or not there is an abnormality in the conversation (step S407). If there is no abnormality, the process returns to step S402, and the conversation time is continuously measured. If there is an abnormality, this is notified to the administrator 103 (step S408), and the process is terminated.

  The CPU 201 performs processing such as calculation of the conversation rate, calculation of the average value, comparison of the conversation information, and determination of whether or not there is an abnormality in the conversation based on the conversation information temporarily stored in the RAM 202. It is executed by doing.

  Here, the notification to the administrator 103 when it is determined that there is an abnormality will be described. FIG. 7 is a diagram illustrating an example of a notification screen to the administrator when it is determined that the conversation is abnormal. In FIG. 7, a plurality of buttons 710 are arranged. Each button 710 is associated with a seat of the operator 102, and a button 710 a corresponding to the operator 102 who is away is displayed as “separated” and a button corresponding to the operator 102 in which an abnormal conversation occurs. In “710b”, “abnormal” is displayed. The administrator 103 can capture and display the screen viewed by the operator 102 corresponding to the button 710 in the capture area 720 by pressing each button 710.

  The manager 103 can give advice (memo, verbal, chat, etc.) to the operator 102 as necessary when a conversation abnormality is notified, or can act on behalf of the operator 102 in some cases.

  In addition, as a material for determining whether or not there is an abnormality in the conversation, the ratio of the time spoken over the entire conversation time at the same time may be used as the determination material in addition to the above. When the ratio of uttering at the same time is large, it is considered that the operator 102 and / or the customer 101 wants to strongly assert his / her opinion. When the customer 101 puts a conversation, there is a possibility that a complaint has occurred. Further, when the operator 102 puts a conversation, there is a possibility that the operator 102 forgets the attitude of listening to the talk and sells it unilaterally. As described above, when the ratio of the time to speak over the entire conversation time is large, there is a possibility that an abnormality has occurred in the conversation.

  Further, as a material for determining whether there is an abnormality in the conversation, a single question and answer time (ratio) in the entire conversation may be used as the material for determining. In a normal normal question and answer session, a conversation catch ball is performed in which one side answers the question and the other side answers the answer (detailed explanation of the product description, etc.). Except time). However, even if the conversation ratio of the whole conversation is within a predetermined threshold range, if the conversation ratio is not uniform in one question-and-answer session, one-sided sales and complaints occur. There is a possibility.

  For example, it is assumed that there are five conversation catch balls during a conversation time of 10 minutes. If the conversation ratio is 9: 1 for the first two times and the ratio for the subsequent three conversations is 2: 8, the conversation ratio for the entire conversation will be about 5: 5. At first glance, it seems like a normal conversation if you focus only on the results. However, there is a possibility that the operator 102 makes a unilateral push sale in the first two times, and the customer 101 makes a complaint about the push sale in the next three times. Even in such a case, it is determined that the abnormality is a conversation. The timing for determining an abnormality varies depending on the business and environment, but it is determined to be abnormal when a non-uniform question-and-answer session continues for a certain number of times (for example, after the end of the second catch ball in the above example). May be.

  As described above, according to the conversation abnormality detection device according to the present embodiment, the conversation time between the operator and the other party is measured, and the conversation information analyzed based on the ratio of the respective conversation times and the preset reference The information is compared to determine whether there is an abnormality in the conversation. Therefore, it is possible to accurately and easily detect abnormalities in conversation depending on the conversation time, which is a formal element, without being influenced by the characteristics of the operator's voice or intentional changes in voice.

  In addition, since the reference information can be set in advance for each business, it is possible to handle conversations in various business, and it is possible to accurately detect by utilizing the characteristics of each business about the rate of conversation. it can.

  Furthermore, an average value of conversation information of a plurality of operators is calculated, and an abnormality is detected in the conversation of the operator whose conversation information exceeds the predetermined threshold range from the average value. Therefore, it becomes possible to compare the whole and the individual, and it is possible to reliably detect an abnormality in the conversation that protrudes from the whole.

Furthermore, in order to analyze the conversation information based on the conversation time required for one question and answer in the conversation between the operator and the other party, based on one or more questions and answers rather than a ratio in the whole conversation. By detecting the abnormality, the abnormality can be detected more accurately.
Furthermore, since the analysis is performed based on the conversation time required for one question-and-answer session, an abnormality can be detected at an early stage when an abnormality has occurred at an early stage of the conversation.

  Furthermore, it is determined that there is an abnormality in the conversation when the simultaneous utterance ratio of the conversation time uttered simultaneously by the operator and the other party with respect to the total conversation time exceeds a predetermined threshold. Since the simultaneous occurrence ratio is high, it occurs frequently when the other party's conversation is interrupted and he / she talks, so that the abnormality of the conversation can be reliably detected.

  Furthermore, the voice characteristics in the conversation are extracted, and only the voice generated by the operator and / or the other party is measured as the conversation time. Therefore, other voices such as a holding tone can be measured by removing from the conversation time, and the analysis can be performed based on the accurate conversation time.

(Second embodiment of the present invention)
The conversation abnormality detection apparatus according to the present embodiment will be described with reference to FIGS. In the present embodiment, descriptions overlapping with those in the first embodiment are omitted.
(1. Configuration)
Below, the structure of the conversation abnormality detection apparatus which concerns on this embodiment is demonstrated.
FIG. 8 is a functional block diagram of the conversation abnormality detection device according to the present embodiment. The conversation abnormality detection device 300 includes a conversation time measurement unit 310, a conversation information analysis unit 320, a conversation abnormality determination unit 330, a reference information unit 340, a voice conversion unit 350, a word detection unit 360, and a conversation time correction unit 370.

  The voice conversion unit 350 performs processing for converting voice conversation between the customer 101 and the operator 102 into character information. Since the technology for converting voice information into character information is a generally known technology, detailed description thereof is omitted.

  The word detection unit 360 performs processing for detecting a predetermined word from the character information converted by the voice conversion unit 350. Predetermined wording is a wording that repeats the same content as the previous sentence, and includes, for example, “That ’s right”, “So it”, “As I said earlier”, “I will repeat” It is. It should be noted that words for delving into the conversation contents such as “specifically” and “details” may not be detected as words for repeating the same contents as the previous sentence.

  When the word detection unit 360 detects a predetermined word, the conversation time correction unit 370 sets the conversation time corresponding to the previous sentence of the word or the conversation time corresponding to the sentence including the word to zero. A process for correcting the overall conversation time is performed.

The conversation information analysis unit 320 analyzes the conversation information based on the conversation time corrected by the conversation time correction unit 370.
Note that the functions of the conversation time measurement unit 310, the conversation information analysis unit 320, the conversation abnormality determination unit 330, and the reference information unit 340 are the same as those in the first embodiment, and a description thereof will be omitted.

(2. Operation)
Below, operation | movement of the conversation abnormality detection apparatus which concerns on this embodiment is demonstrated.
FIG. 9 is a flowchart showing the operation of the conversation abnormality detection device according to the present embodiment. First, reference information is input as input data 301 and stored in the reference information unit 340 (step S901). When the reference information is stored, the conversation time of the operator 102 and the customer 101 is measured (step S902). The conversation voice of the operator 102 and the customer 101 is converted into character information (step S903), and a word indicating repetition is retrieved from the converted character information (step S904). As a result of the search, it is determined whether or not a word indicating repetition is detected (step S905), and if not, the process proceeds to step S907. If detected, the conversation time is corrected by setting the conversation time of the overlapping portion to zero for the conversation of the same content (step S906).

  Here, the correction of the conversation time will be described. FIG. 10 is a diagram showing processing when correcting the conversation time. The voice conversation between the customer 101 and the operator 102 is converted into character information. As described above, the technology for converting voice information into character information is generally a well-known technology, and a detailed description thereof will be omitted. When attention is paid to the conversation of the operator 102, the customer 101 talks after talking "XXXXX" in 30 seconds, and then continues as "XXXXX as I said earlier" for 50 seconds. Have a conversation.

  A predetermined word is detected from the character information of this conversation. The predetermined word is a word indicating repetition and is registered in advance. Here, "Like I've done it earlier" is detected as a repeated wording.

  The fact that a predetermined word has been detected means that the previous sentence and the present sentence including the word in the operator's conversation have the same content, and are overlapping conversations. Therefore, here, the current sentence (50 seconds) including a predetermined word is corrected to 0 seconds. Note that the preceding sentence (30 seconds) may be corrected to 0 seconds. As described above, the conversation time measured by the conversation time measurement unit 310 is corrected based on the conversation contents and analyzed by the conversation information analysis unit 320.

  Returning to FIG. 9, the processing from step S907 to step S912 is the same as the processing from step S403 to step S408 in FIG.

As described above, according to the conversation abnormality detection device according to the present embodiment, conversation voice information is converted into character information, a word for repeating the same content as the previous sentence is detected, and the word is detected. Analyzes the conversation information ignoring the conversation time of the previous sentence or the present sentence. For this reason, conversation information is analyzed except for redundant explanations and unnecessary conversation time when a conversation is performed, and the conversation information can be accurately analyzed to detect an abnormality.
In addition, since there are many redundant conversations, there are many cases where the explanation is not good, and it is easy for such an operator to provide conversation guidance.

  Although the present invention has been described with the above embodiments, the technical scope of the present invention is not limited to the scope described in the embodiments, and various modifications or improvements can be added to these embodiments. . And embodiment which added such a change or improvement is also contained in the technical scope of the present invention. This is apparent from the claims and the means for solving the problems.

(Other embodiments)
The following additional notes will be given with respect to the above embodiments.
(Supplementary Note 1) Reference information storage means for presetting and storing reference information on conversation time between the operator and the other party for each job, conversation time for measuring the conversation time of the operator and the conversation time of the other party Measuring means; conversation information analyzing means for analyzing conversation information between the operator and the other party based on a ratio of each conversation time measured by the conversation time measuring means; conversation information analyzed by the conversation information analyzing means; A conversation abnormality detection device comprising abnormality determination means for comparing the reference information stored in the reference information storage means and determining whether or not there is an abnormality in the conversation.

  (Supplementary note 2) In the conversation abnormality detecting device according to supplementary note 1, a conversion means for converting conversation voice information into character information, and a wording for repeating the same contents as the preceding sentence from the character information converted by the conversion means. When the word detection means to detect and the word detection means detect a word for repeating the same contents as the previous sentence, the conversation time of the previous sentence of the relevant word or the present sentence including the relevant word is ignored. A conversation time correction unit that corrects the conversation time, and the conversation information analysis unit analyzes the conversation information based on the conversation time corrected by the conversation time correction unit.

  (Supplementary Note 3) In the conversation abnormality detecting device according to Supplementary Note 1 or 2, there are a plurality of the operators, and the conversation information analysis unit calculates an average value of all operators for the analyzed conversation information, and the abnormality determination unit However, when conversation information exceeding a predetermined threshold range is detected from the average value calculated by the conversation information analysis means, the conversation abnormality detection device determines that the conversation is abnormal for the conversation information.

  (Supplementary note 4) In the conversation abnormality detecting device according to any one of supplementary notes 1 to 3, the conversation information analyzing unit calculates a simultaneous utterance ratio of conversation time in which the operator and the other party uttered simultaneously with respect to a total conversation time. A conversation abnormality detection device, wherein the abnormality determination unit determines that there is an abnormality in the conversation when the calculated simultaneous utterance ratio exceeds a predetermined threshold.

  (Appendix 5) In the conversation abnormality detecting device according to any one of appendices 1 to 4, the conversation information analyzing means is based on a conversation time required for one question and answer in a conversation between the operator and the other party, Conversation abnormality detection device that analyzes conversation information.

  (Supplementary note 6) The conversation abnormality detecting device according to any one of supplementary notes 1 to 5, further comprising feature extraction means for extracting a feature of speech in conversation, wherein the speech time measurement means extracts the voice extracted by the feature extraction means The conversation abnormality detection apparatus which measures only the voice which the operator and / or the other party of the call generate | occur | produced as the conversation time from the characteristic of this.

  (Appendix 7) A conversation time measurement step for measuring the conversation time of the operator and the conversation time of the other party, and the ratio between the conversation time measured in the conversation time measurement step and the operator and the other party The conversation information analysis step for analyzing the conversation information, the conversation information analyzed in the conversation information analysis step, and the reference information for each business of the operator and the other party of communication set in advance are compared, and the conversation is abnormal. A method for detecting abnormalities in conversation, comprising: an abnormality determining step for determining whether or not.

  (Supplementary note 8) In the conversation abnormality detection method according to supplementary note 7, a conversion step for converting conversation voice information into character information, and a wording for repeating the same content as the previous sentence from the character information converted in the conversion step When a word for detecting the same content as the preceding sentence is detected in the word detecting step, the conversation time of the sentence before the sentence or the sentence including the word is detected. Conversation time correction step of correcting conversation time by ignoring the conversation time, and in the conversation information analysis step, conversation information is analyzed based on the conversation time corrected in the conversation time correction step .

  (Supplementary note 9) In the conversation abnormality detection method according to supplementary note 7 or 8, there are a plurality of the operators, and an average value of all operators is calculated for the conversation information analyzed in the conversation information analysis step, and the abnormality determination A conversation abnormality detection method in which, when conversation information exceeding a predetermined threshold range from the average value calculated in the conversation information analysis step is detected in step, the conversation abnormality is determined to be abnormal for the conversation information .

  (Supplementary note 10) In the conversation abnormality detection method according to any one of supplementary notes 7 to 9, in the conversation information analysis step, the simultaneous utterance ratio of the conversation time in which the operator and the call partner uttered simultaneously with respect to the total conversation time is calculated. In the abnormality determination step, when the calculated simultaneous utterance ratio exceeds a predetermined threshold, it is determined that there is an abnormality in the conversation.

  (Supplementary note 11) In the conversation abnormality detection method according to any one of supplementary notes 7 to 10, in the conversation information analysis step, based on a conversation time required for one question and answer in a conversation between the operator and the other party, Conversation abnormality detection method that analyzes conversation information.

  (Supplementary note 12) In the conversation abnormality detection method according to any one of supplementary notes 7 to 11, including the feature extraction step of extracting speech features in conversation, the speech extracted in the feature extraction step in the conversation time measurement step The conversation abnormality detection method in which only the voice generated by the operator and / or the communication partner is measured as the conversation time.

  (Supplementary note 13) Reference information storage means for presetting and storing reference information on conversation time between the operator and the other party for each job, conversation time measurement for measuring the conversation time of the operator and the conversation time of the other party Means, conversation information analysis means for analyzing conversation information between the operator and the other party based on the ratio of each conversation time measured by the conversation time measurement means, conversation information analyzed by the conversation information analysis means, and the reference information A conversation abnormality detection program that causes a computer to function as an abnormality determination unit that compares reference information stored in a storage unit to determine whether or not there is an abnormality in a conversation.

  (Supplementary note 14) In the conversation abnormality detection program according to supplementary note 13, conversion means for converting conversational voice information into character information, and wording for repeating the same content as the previous sentence is detected from the character information converted by the conversion means When the word detection means detects a word for repeating the same content as the previous sentence, the conversation time of the sentence before the sentence or the sentence including the word is ignored. A conversation abnormality detection program for causing a computer to function as conversation time correction means for correcting conversation time, wherein the conversation information analysis means analyzes conversation information based on the conversation time corrected by the conversation time correction means.

  (Supplementary note 15) In the conversation abnormality detection program according to supplementary note 13 or 14, there are a plurality of the operators, and the conversation information analysis unit calculates an average value of all operators for the analyzed conversation information, and the abnormality determination unit However, a conversation abnormality detection program for determining that there is an abnormality in the conversation when the conversation information exceeding a predetermined threshold range is detected from the average value calculated by the conversation information analysis means.

  (Supplementary note 16) In the conversation abnormality detection program according to any one of supplementary notes 13 to 15, the conversation information analysis unit calculates a simultaneous utterance ratio of conversation time in which the operator and the other party speak simultaneously with respect to a total conversation time. A conversation abnormality detection program for determining that the conversation is abnormal when the calculated simultaneous utterance ratio exceeds a predetermined threshold.

  (Supplementary note 17) In the conversation abnormality detection program according to any one of supplementary notes 13 to 16, the conversation information analyzing means is based on a conversation time required for one question and answer in a conversation between the operator and the other party. A conversation anomaly detection program that analyzes conversation information.

  (Supplementary note 18) In the conversation abnormality detection program according to any one of supplementary notes 13 to 17, the computer is caused to function as feature extraction means for extracting voice features in conversation, and the conversation time measurement means is A conversation abnormality detection program for measuring only voices generated by an operator and / or a communication partner as conversation time from extracted voice characteristics.

DESCRIPTION OF SYMBOLS 100 Operator management system 101 Customer 102 Operator 103 Manager 200 Computer 201 CPU
202 RAM
203 ROM
204 Flash memory 205 HD
206 LAN Card 207 Mouse 208 Keyboard 209 Video Card 209a Display 210 Sound Card 210a Speaker / Microphone 211 Drive 300 Conversation Abnormality Detection Device 301 Input Data 310 Conversation Time Measurement Unit 320 Conversation Information Analysis Unit 321 Conversation Ratio Calculation Unit 322 Average Value Calculation Unit 330 Conversation abnormality determination unit 340 Reference information unit 350 Audio conversion unit 360 Word detection unit 370 Conversation time correction unit 710 button 710a Away button 710b Abnormal button 720 Capture region

Claims (12)

Conversation time measuring means for measuring the utterance time of the operator and the utterance time of the operator's call partner,
Based on each utterance time measured by the conversation time measuring means, a conversation information analyzing means for calculating a ratio of utterance by the operator, or a ratio of utterance by the other party,
A conversation anomaly detection apparatus comprising: means for outputting a display corresponding to a result of comparing the ratio calculated by the conversation information analysis means and the ratio of the utterance to be compared. In the conversation abnormality detection device according to claim 1,
Conversion means for converting voice information included in utterances by the operator or the other party into character information;
A word detection means for detecting a word for repeating the same contents as the contents shown in the character information converted in the past by the conversion means from the character information converted by the conversion means;
When the word detection means detects the word, a conversation for correcting the utterance time of the character information converted in the past showing the same content as the word or the character information converted by the conversion means including the word Time correction means,
The conversation abnormality detecting device, wherein the conversation information analyzing means calculates the ratio based on the utterance time corrected by the conversation time correcting means. In the conversation abnormality detection device according to claim 1 or 2,
The comparison-target utterance ratio is a pre-stored reference utterance ratio or an average value of utterance ratios calculated from a plurality of operator utterance ratios calculated by the conversation information analysis means. ,
The output means outputs information indicating that there is an abnormality in the conversation for the operator when a ratio exceeding a predetermined threshold range from the average value calculated by the conversation information analysis means is detected. Conversation abnormality detection device characterized by that. In the conversation abnormality detection device according to any one of claims 1 to 3,
The conversation information analyzing means calculates a simultaneous utterance ratio that is a ratio of a time when the operator and the call partner uttered simultaneously with respect to a total conversation time, and the outputting means outputs information on the simultaneous utterance ratio. Conversation abnormality detection device characterized by that. Computer
A conversation time measuring step for measuring the utterance time of the operator and the utterance time of the operator's call partner,
Based on each utterance time measured in the conversation time measurement step, a conversation information analysis step of calculating a ratio of utterance by the operator or a ratio of utterance by the other party,
A method for detecting an abnormality in conversation, comprising: executing a step of outputting a display according to a result of comparing the ratio calculated in the conversation information analyzing step and the ratio of the utterance to be compared. In the conversation abnormality detection method according to claim 5,
Computer
A conversion step of converting voice information included in utterances by the operator or the other party into character information;
A word detection step for detecting a word for repeating the same content as the content shown in the character information converted in the past by the conversion step, from the character information converted in the conversion step;
When the word is detected in the word detection step, the utterance time of the character information converted in the past showing the same content as the word or the character information converted by the conversion step including the word is corrected. And a conversation time correction step,
The conversation abnormality detecting method, wherein the conversation information analyzing step calculates the ratio based on the utterance time corrected in the conversation time correcting step. In the conversation abnormality detection method according to claim 5 or 6,
The comparison-target utterance ratio is a pre-stored reference utterance ratio or an average value of utterance ratios calculated from a plurality of operator utterance ratios calculated in the conversation information analysis step. ,
The outputting step outputs information indicating that there is an abnormality in the conversation for the operator when a ratio exceeding a predetermined threshold range is detected from the average value calculated in the conversation information analyzing step. Conversation abnormality detection method characterized by this. The conversation abnormality detection method according to any one of claims 5 to 7,
The conversation information analyzing step calculates a simultaneous utterance ratio that is a ratio of a time when the operator and the other party uttered simultaneously with respect to a total conversation time, and the outputting step outputs information on the simultaneous utterance ratio. Conversation abnormality detection method characterized by this. On the computer,
A conversation time measuring step for measuring the utterance time of the operator and the utterance time of the operator's call partner,
Based on each utterance time measured in the conversation time measurement step, a conversation information analysis step of calculating a ratio of utterance by the operator or a ratio of utterance by the other party,
Outputting a display according to the result of comparing the ratio calculated in the conversation information analysis step and the ratio of the utterance of the comparison target;
A conversation abnormality detection program characterized by causing In the conversation abnormality detection program according to claim 9,
A conversion step of converting voice information included in utterances by the operator or the other party into character information;
A word detection step for detecting a word for repeating the same content as the content shown in the character information converted in the past by the conversion step, from the character information converted by the conversion step;
When the word detection step detects the word, the conversation that corrects the utterance time of the character information converted in the past showing the same content as the word or the character information converted by the conversion step including the word Further causing the computer to perform a time correction step;
The conversation abnormality detecting program characterized in that the conversation information analyzing step calculates the ratio based on the utterance time corrected by the conversation time correcting means. In the conversation abnormality detection program according to claim 9 or 10,
The comparison-target utterance ratio is a pre-stored reference utterance ratio or an average value of utterance ratios calculated from a plurality of operator utterance ratios calculated in the conversation information analysis step. ,
The outputting step outputs information indicating that there is an abnormality in the conversation for the operator when a ratio exceeding a predetermined threshold range is detected from the average value calculated in the conversation information analyzing step. Conversation abnormality detection program characterized by that. In the conversation abnormality detection program according to any one of claims 9 to 11,
The conversation information analyzing step calculates a simultaneous utterance ratio that is a ratio of a time when the operator and the other party uttered simultaneously with respect to a total conversation time, and the outputting step outputs information on the simultaneous utterance ratio. Conversation abnormality detection program characterized by that.

Images