JP2007286097A - Voice reception claim detection method and device, and voice reception claim detection program and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To promptly detect the generation of a claim and to smoothly process the claim in reception work by voice. <P>SOLUTION: A voice reception claim detection method includes a reception start time detection process for detecting reception start time from the utterance level of key input or voice input; a response utterance start time detection process for detecting the response utterance start time; a claim detection evaluation value measurement process for detecting response time which is the difference of response utterance start time and reception start time, as a claim detection evaluation value; a claim detection process for determining it to be claim processing, if the claim detection evaluation value is less than a threshold; and an output processing for outputting operator information under processing, when it is determined as being a claim in the claim detection processing. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a voice reception complaint detection method suitable for application to a voice reception system via a telephone or the Internet, and more particularly, a voice reception complaint detection method, apparatus, program, and recording medium that allow a call center administrator to determine a customer's condition. About.

As a voice reception complaint detection device, for example, as disclosed in Patent Document 1, the words, speed, and volume spoken by a customer are measured, and compared with the normal customer's state, the customer's emotional state is determined from the customer's voice. There are techniques to estimate.
JP 2004-37989 A

The conventional techniques have the following problems.
(1) It is difficult to detect the words spoken by the customer. ・ Voice recognition performance is due to the customer's voice being distorted by the telephone line, the influence of background noise on the customer side, the wraparound on the operator side, etc. Can't identify words that speak badly.
-When a customer is angry, it is difficult to prepare words to be detected in advance for voice recognition because it is not known what words are spoken.
(2) It is difficult to detect the customer's condition based on the customer's speaking speed. Particularly, the excitement does not necessarily increase the speaking speed.
・ Even if excitement increases the speaking speed, it is difficult to calculate the speaking speed.
(3) Volume measurement accuracy is not high ・ Volume varies depending on the distance from the customer's phone terminal and customer's mouth to the microphone of the phone terminal, so the degree of excitement cannot be estimated by volume alone.

  In order to solve these problems, the voice reception claim detection method according to the present invention includes a reception start time detection process for detecting a reception start time from a key input or an utterance level, and a conflict utterance start time detection process for detecting a conflict utterance start time. And a claim detection evaluation value measurement process for detecting a response time obtained as a difference between the start time of the utterance and the reception start time as a claim detection evaluation value, and a claim for determining that the claim is handled if the claim detection evaluation value is equal to or less than a threshold value It includes a detection process and an output process for outputting operator information being received when it is determined as a complaint in the complaint detection process.

In the present invention, further recognition of input speech is performed to detect a pre-registered conflict word, to obtain a start time and an end time of the conflict word, and the number of syllables or phonemes of the registered conflict word Claim detection evaluation value measurement processing for measuring the speech utterance speed obtained by dividing the word by the word duration of the complaint as a claim detection evaluation value, and claim detection processing for determining that the claim is handled if the claim detection evaluation value is equal to or greater than a threshold value And an output process for outputting the operator information being received when it is determined that the complaint is detected in the complaint detection process.
In the present invention, recognition processing of input speech is further performed to detect a pre-registered companion word, and a companion utterance recognition process for obtaining a start time and an end time of the companion word; Volume measurement processing for measuring the volume, claim detection evaluation value measurement processing for measuring the volume of the utterance utterance by dividing the volume of the utterance utterance by the duration of the utterance word as a complaint detection evaluation value, and the volume of the utterance utterance volume And claim output processing for outputting operator information during response when it is determined as a complaint in the claim detection processing.

  In the present invention, the input speech recognition process is further performed to detect a pre-registered conflict word, and to obtain a start time and an end time of the conflict word, and from the conflict utterance start time, Measures the relative speech interval measurement process to measure the relative speech interval obtained by subtracting the end time of the previous relative speech, and measures the average value of the relative speech intervals generated during a certain time as the claim detection evaluation value Claim detection evaluation value measurement processing to be performed, claim detection processing to determine that a complaint is handled if the complaint detection evaluation value is equal to or less than a threshold value, and output to output operator information during response when the complaint detection unit determines a complaint Processing.

The present invention further includes an integrated claim detection process for detecting a claim using an integrated claim evaluation value obtained by integrating two or more of the claim detection evaluation values with a weighted linear sum.
In the present invention, the integrated claim detection that determines a claim when at least two or more of the claim detection evaluation values of the response time of the input speech, the relative speech speed, the relative speech volume level, and the relative speech interval are determined to be a complaint. Including processing.
The present invention further includes a claim detection process in which a claim detection evaluation value or an integrated claim evaluation value and a threshold value are normalized by an average value for each input voice type and detected.

The present invention further includes a detection parameter update process for returning the average value of the claim detection evaluation value or the integrated claim evaluation value and the detection threshold value to the reference value after a lapse of a predetermined time.
The voice reception claim detection device according to the present invention includes a reception start time detection unit that detects a reception start time from a key input or a speech level, a conflict speech start time detection unit that detects a conflict speech start time, a conflict speech start time, and a speech start A claim detection evaluation value measuring unit that detects a response time obtained as a time difference as a claim detection evaluation value, a claim detection unit that determines that a claim is handled if the claim detection evaluation value is equal to or less than a threshold value, and a claim detection process And an output unit that outputs the operator information being received when it is determined.

  The speech reception claim detection apparatus according to the present invention recognizes input speech, detects a pre-registered conflict word, obtains a start time and end time of the conflict word, and the registered conflict word A complaint detection evaluation value measuring unit that measures the speech utterance speed obtained by dividing the number of syllables or phonemes by the word duration of complaints as a complaint detection evaluation value; A complaint detection unit, and an output unit that outputs operator information in response to a complaint detected by the complaint detection unit.

  The voice reception claim detection apparatus according to the present invention further performs a process of recognizing input voice, detects a pre-registered conflict word, obtains a start time and an end time of the conflict word, A volume measurement unit that measures the volume of the utterance utterance during word recording, and a claim detection evaluation value measurement that measures the utterance volume level obtained by dividing the volume of the utterance utterance by the duration of the utterance word as a claim detection evaluation value. A claim detection unit that determines that a complaint is handled if the complaint detection evaluation value is equal to or greater than a threshold value, and an output unit that outputs operator information during response when the complaint detection unit determines that the complaint is a complaint. Features.

  The voice reception claim detection apparatus according to the present invention performs a process for recognizing input voice, detects a pre-registered conflict word, obtains a start time and an end time of the conflict word, and The average value of the relative utterance intervals of the relative utterances generated during a certain period of time and the relative utterance interval measurement unit that measures the relative utterance interval obtained by subtracting the end time of the previous utterance utterance from the utterance start time. A claim detection evaluation value measuring unit that measures as a claim detection evaluation value, a claim detection unit that determines that a claim is handled if the claim detection evaluation value is less than or equal to a threshold value, and a response that is being handled when the claim detection unit determines a claim And an output unit for outputting operator information.

The voice reception claim detection apparatus according to the present invention includes an integrated claim detection unit that detects a claim using an integrated claim evaluation value obtained by integrating at least two of the claim detection evaluation values with a weighted linear sum.
The voice reception claim detection apparatus according to the present invention further includes an integrated claim detection unit that determines a claim when at least two of the claim detection evaluation values are determined to be a claim.
The voice reception claim detection apparatus according to the present invention further includes a claim detection unit that performs detection by normalizing the claim detection evaluation value or the integrated claim evaluation value and their threshold values with an average value for each input voice type. To do.

The voice reception claim detection apparatus according to the present invention further includes a detection parameter update unit that returns the average value of the claim detection evaluation value or the integrated claim evaluation value and the detection threshold value to a reference value after a predetermined time has elapsed.
The voice reception claim detection program according to the present invention is described in a computer-readable program language, and causes the computer to execute any one of the voice reception claim detection methods described above.
The recording medium on which the voice reception claim detection program according to the present invention is recorded is constituted by a computer-readable recording medium, and the voice reception claim detection program is recorded on the recording medium.

In general, whether the customer is excited or not is judged from the voice quality of the customer immediately after receiving the call and shows that the customer is responding properly. Also, attention is paid to the fact that the response utterance speed at this time is faster than normal, and that the response utterance such as the operator's interaction is faster and the volume is higher. Further, attention is paid to the fact that the interval between competing utterances of the operator receiving the complaint is shorter than the normal interval.
For this reason, in the present invention, the detection target is limited to the operator voice, and “the length of time from the time immediately after receiving the call until the operator speaks (response time after receiving the call)” and “the speed of the talk utterance speed (the answer utterance speed)” And “the volume of the voice of the sumo utterance (the sumo utterance volume)” and “the interval of the sumo utterance (the sumo utterance interval)”, it is detected whether or not the customer who is responding to the operator is in an excited state. The operator (supervisor) is notified, and it becomes possible to help the hearing and the operator and respond smoothly.

Furthermore, since the detection target is limited to the operator's voice, since the companion word spoken by the operator is a typical word, the detection target vocabulary can be easily selected. In addition, the operator's voice has high sound quality and a high recognition rate, so that the detection performance is high. In other words, since the range of the average voice volume level is limited after the operator wears the headset, the detection performance based on the volume is high.
In addition, since detection is possible with the utterances immediately after the reception, the time required for the detection process is small. In addition, the past “response time after reception”, “conflict utterance speed”, “conflict utterance volume”, and “interference utterance interval” are saved for each operator, and the detection performance is further improved by normalizing them with the average value. I can do it.

  Therefore, according to the present invention, the customer excitement (claim) detection performance is higher than that of the prior art. In addition, it is easy to register a companion word for excitement (claim) detection. Furthermore, the time required for detecting excitement (claim) can be shortened.

The voice reception claim detection method and apparatus according to the present invention can be realized entirely by hardware, but in order to realize the simplest, the voice reception claim detection program proposed in the present invention is installed in a computer, and the computer The embodiment that allows the voice reception complaint detection apparatus to execute the voice reception complaint detection method according to the present invention is the best.
In order for a computer to function as a voice reception complaint detection device according to the present invention, a voice reception complaint detection program installed in a computer detects a reception start time from a key input or voice input speech level by using a voice reception complaint detection program; A conflict utterance start time detection process for detecting a start time, a claim detection evaluation value measuring unit for detecting a response time obtained from a difference between the conflict utterance start time and the reception start time as a claim detection evaluation value, and the claim detection evaluation value is a threshold value As a voice reception claim device, a complaint detection unit that determines a response to a complaint if it is below, and an output unit that outputs operator information during response when the complaint detection unit determines “complaint”. An embodiment to function;
It is obtained by recognizing the input speech, detecting the conflicting word and obtaining the start time and end time of the conflicting word, and dividing the number of syllables or phonemes of the conflicting word by the duration of the conflicting word. A complaint detection evaluation value measuring unit that measures as a conflicting speech rate complaint detection evaluation value, a claim detection unit that determines a response to a “claim” according to the measurement result of the claim detection evaluation value measurement unit, and a “claim” An embodiment for constructing an output unit that outputs operator information in response to a case, and functioning as a voice reception device;
It is obtained by detecting the companion word and obtaining the start time and end time of the companion word, the volume measuring unit for measuring the volume of the companion utterance, and dividing the volume of the companion utterance by the companion word duration. A complaint detection evaluation value measurement unit that measures the volume level of the utterance utterance as a complaint detection evaluation value, a complaint detection unit that determines the response of the “claim” according to the measurement result of the complaint detection evaluation value measurement unit, and a determination as “claim” An output unit that outputs operator information in response to the response, and an embodiment for functioning as a voice reception device;
A companion utterance recognition unit that detects a companion word from the input speech and obtains the start time and end time of the companion word, and a subtraction obtained by subtracting the end time of the previous companion utterance from the start of the companion utterance Conflict utterance interval measurement unit that measures the utterance interval, complaint detection evaluation value measurement unit that measures the average value of the relative utterance intervals of the conflicts that occur during a certain time as a claim detection evaluation value, and measurement results of the claim detection evaluation value An embodiment for constructing a complaint detection unit that determines a response of a “claim” according to a response, and an output unit that outputs operator information that is being answered when it is determined to be a “claim”, and functions as a voice reception complaint detection device You could think so.

In the first embodiment, a method for determining whether or not a complaint is made using a length of time from immediately after receiving a call until the operator starts speaking (response time after receiving RT: Response Time) will be described.
A functional configuration diagram of the first embodiment is shown in FIG. 1, and a processing procedure is shown in FIG.
In the first embodiment, the reception start time detection unit 13 determines the reception start time from the level of the operator or user voice input to the audio input terminal 12, the operator's keyboard or mouse input given to the input terminal 11, or the reception operation signal. It detects (step SP21, SP22). Next, the claim detection evaluation using the response time RT (Response Time) after reception obtained by the difference between the start time and the reception start time of the operator from the voice of the operator (“interaction speech start time−reception start time”) as the claim detection evaluation value The value is measured by the value measuring unit 15 (step SP24).

If the claim detection evaluation value obtained by the response time RT is equal to or less than the threshold value, the claim detection unit 16 determines that the response is a complaint (step SP25). If it is determined that a complaint is received, the operator information during the response is output from the output unit 17. For example, the output operator information is transmitted from the output unit 17 to the supervisor (call center) (step SP26).
Further, the average value of the response time RT of the operator is stored in the response time average memory (see the claim detection evaluation value average memory 111 in FIG. 11), and the response time RT and the detection response time threshold are normalized by the response time average. As a result, robustness of detection accuracy can be improved.

  In the utterance start time detection unit 14, the operator / user voice separation unit 18 first separates the operator voice and the user voice by using hardware such as an existing separation adapter or a sound source separation technique (step SP27), and then the operator voice. A method based on operator / user voice separation that detects the speech level at the voice level is conceivable. Further, even when the operator voice and the user voice are not separated, the talk utterance recognition unit 19 detects the voice talk words specific to the operator that are assumed in advance (for example, “Yes”, “Yes”, “Sorry”) by voice recognition. Thus, a method based on voice recognition (step SP28) for detecting operator-interactive speech is also possible. The detection accuracy can be further improved by combining both the method based on operator / user speech separation and the method based on speech recognition.

  In addition, since the method based on speech recognition is already known by the speaker (operator), the detection performance can be further improved by tuning the speech or language model used for speech recognition for a specific speaker for each speaker. It becomes possible.

In the second embodiment, it is determined whether or not a complaint is made by using the speed of speaking speech (Speech Rate SR).
FIG. 3 shows a functional configuration diagram of the second embodiment, and FIG. 4 shows a processing procedure.
In the second embodiment, the conflicting speech recognition unit 19 first detects the conflicting word from the operator's voice. The companion utterance recognition unit 19 recognizes the operator's voice, detects a preregistered companion word, and obtains the start time and end time of the companion word (step SP41). Next, the complaint detection evaluation value measurement unit 15 divides the number of syllables or phonemes of the registered reference word by the reference word duration (“completion word end time−start time”) to obtain the reference speech speed. SR is measured as a claim detection evaluation value (step SP42). If the competing speech rate SR obtained as the complaint detection evaluation value is equal to or higher than the threshold, the complaint detection unit 16 determines that the complaint is handled (step SP43). When it is determined that the complaint is received, the output unit 17 outputs the operator information being handled. The operator information is transmitted to the supervisor (step SP44).

In addition, the average value of the operator utterance speed is stored in a memory (see FIG. 11), and the robustness of the detection accuracy can be improved by normalizing the utterance speed and the complaint detection utterance speed threshold with the utterance speed average. . The utterance speed average is not limited to the conflicting utterance, but can be normalized more stably by detecting it including the normal operator utterance.
As in the first embodiment, the process of the interaction speech recognition unit 19 can improve accuracy by combining methods based on the operations of the reception start time detection unit 13 and the operator / user voice separation unit 18 (steps SP45 and SP46). In addition, the process of the interaction speech recognition unit 19 can be tuned for the operator.

  Furthermore, by detecting the reception start time (step SP45), the recognition target (step SP41) after the reception start time is used as the recognition target (step SP41), so that the accuracy of recognizing the utterance can be improved.

In the third embodiment, it is determined whether or not the claim is a complaint using the volume of the voice of the companion utterance (interference utterance volume level PL: Power Level).
FIG. 5 shows a functional configuration example of the third embodiment, and FIG. 6 shows a processing procedure.
In the third embodiment, first, the conflicting speech recognition unit 19 detects the conflicting word from the operator's voice (step SP61). The companion utterance recognition unit 19 recognizes the operator's voice, detects a preregistered companion word, and obtains the start time and end time of the companion word. At the same time, the volume utterance volume level PL at the time of recording the conflicting word is measured by the volume measuring unit 20 (step SP61).

Next, the claim detection evaluation value measuring unit 15 performs claim detection and evaluation on the companion utterance volume level PL obtained by dividing the companion utterance volume level PL by the companion word duration (“completion word end time−start time”). Measured as a value (step SP62).
If the competing speech volume level PL obtained as the complaint detection evaluation value is equal to or greater than the threshold value, the complaint detection unit 16 determines that the complaint is handled (step SP63). If it is determined that the response is a complaint, the output unit 17 transmits the operator information being received to the supervisor (step SP64).

Further, the average value of the operator utterance volume is stored in a memory (refer to the average value memory 111 in FIG. 11), and the robustness of the detection accuracy is improved by normalizing the utterance volume and the detection utterance volume threshold with the utterance volume average. Can be increased. The utterance volume average can be normalized more stably by detecting the utterance volume average including not only the utterance utterance but also the normal operator utterance.
Furthermore, the influence of the headset mounting position can be reduced by changing the threshold value of the operator utterance volume to the default value every time the operator's absence input signal or the attachment / detachment of the headset is detected.

  As in the first embodiment, the processing of the operator's conflict speech recognition unit 19 can be combined with methods based on the operations of the reception start time detection unit 13 and the operator / user voice separation unit 18 to improve accuracy (steps SP65 and SP66). In addition, the process of the interaction speech recognition unit 19 can be tuned for the operator.

In the fourth embodiment, it is determined whether or not it is a complaint by using the interval of the interaction utterance (interaction utterance interval RI: Response Interval).
FIG. 7 shows a functional configuration example of the fourth embodiment, and FIG. 8 shows a processing procedure.
In the fourth embodiment, the conflicting utterance recognition unit 19 first detects a conflicting word from the operator's voice. The companion utterance recognition unit 19 recognizes the operator's voice, detects a preregistered companion word, and obtains the start time and end time of the companion word (step SP81).
The interaction utterance interval measurement unit 21 measures the interaction utterance interval RI of the interaction for each occurrence of the interaction using the start time and the end time of the interaction word (step SP82).

Next, the claim detection evaluation value measurement unit 15 measures the average value of the conflict utterance intervals RI of the conflicts generated during a predetermined time as the claim detection evaluation value (step SP83). If the average value of the conflicting speech intervals, which is the complaint detection evaluation value, is equal to or less than the threshold value, the complaint detection unit 16 determines that the complaint is handled (step SP84). Then, when it is determined that the response is a complaint, the output unit 17 transmits the operator information being received to the supervisor (step SP85).
Similar to the first embodiment, the process of the utterance utterance recognition unit 19 can improve the accuracy by combining methods based on the operations of the reception start time detection unit 13 and the operator / user voice separation unit 18 (steps SP86 and 87). In addition, the process of the interaction speech recognition unit 19 can be tuned for the operator.

In the fifth embodiment, a configuration is proposed in which an integrated complaint detecting unit that integrates the measurement results of the complaint detection evaluation value measuring unit individually handled in the first to fourth embodiments is provided.
FIG. 9 shows a functional configuration example of the fifth embodiment, and FIG. 10 shows a processing procedure.
In the fifth embodiment, the claim detection evaluation value measurement unit 15 used in the first embodiment is referred to as a first claim detection evaluation value measurement unit 15-1, and here the response time RT is measured (step SP100-1).
The claim detection evaluation value measurement unit 15 used in the second embodiment is referred to as a second claim detection evaluation value measurement unit 15-2, and here, the conflicting speech rate SR is measured (step SP100-2).

The claim detection evaluation value measurement unit 15 used in the third embodiment is referred to as a third claim detection evaluation value measurement unit 15-3, and here, the conflicting speech volume PL is measured (step SP100-3).
The claim detection evaluation value measurement unit 15 used in the fourth embodiment is referred to as a fourth claim detection evaluation value measurement unit 15-4, and here, the conflicting speech interval RI is measured (step SP100-4).
The measurement values of the first claim detection evaluation value measurement unit 15-1 to the fourth claim detection evaluation value measurement unit 15-4 are input to the integrated complaint detection unit 91. In the integrated complaint detection unit 91, the response time after reception RT, the relative speech speed SR, the relative speech volume level PL, which are measured by the first complaint detection evaluation value measurement unit 15-1 to the fourth complaint detection evaluation value measurement unit 15-4, Integrated complaint evaluation value obtained by integrating at least two of the mutual speech intervals RI with a weighted linear sum (for example, V = w1 * (1 / RT) + w2 * (SR) + w3 * (PL) + w4 * (1 / RI), wi calculates the weight, i = 1, 2, 3, 4), and if this integrated claim evaluation value V is equal to or greater than the integrated claim evaluation threshold, it is determined as a claim response (step SP101) . Then, when it is determined that the response is a complaint response, the operator information being received by the output unit 17 is transmitted to the supervisor (step SP102).

  In the integrated complaint evaluation, the number of detected claims may be, for example, two or more among the detected evaluation values of the response time RT after reception, the utterance speed SR, and the utterance volume level PL in addition to the weighted linear talk.

In the sixth embodiment, the response time after reception RT, the utterance utterance speed SR, the utterance utterance volume level PL, the utterance utterance interval RI, and the integrated claim evaluation value (hereinafter collectively referred to as the claims) handled in the first to fifth embodiments. An embodiment in which detection accuracy is improved by tuning or resetting the detection evaluation value for each speaker will be described.
An example of the functional configuration of the sixth embodiment is shown in FIG.
In the sixth embodiment, the process up to the measurement of the complaint detection evaluation value in the complaint detection evaluation value measuring unit 15 is the same as that of the first to fifth embodiments. The average of the complaint detection evaluation values is stored in the complaint detection evaluation value average memory 111. The claim evaluation value normalization unit 112 normalizes the claim detection evaluation value and the claim detection threshold for each speaker, that is, for each operator, using the average value of the claim detection evaluation values stored in the claim detection evaluation value average memory 111.

By this normalization, the claim detection evaluation value and the claim detection threshold value used in the claim detection unit 16 are leveled, and generation of values that are significantly different from each other is prevented, thereby improving the claim detection accuracy.
Further, the complaint detection evaluation value average updating unit 113 uses the detection signal of the absence detection unit 114 or the elapsed time measurement unit 116 to return the complaint detection evaluation value average to the default value when the operator leaves the seat or every time a fixed time elapses. (Reset) reduces detection accuracy degradation. An operator's absence detection can be performed by detecting a no-input state of the operator's keyboard or button operation for a certain period of time.

The method and apparatus for detecting a voice reception claim according to the present invention described above installs a voice reception claim detection program for operating a computer in accordance with the processing procedure described in each embodiment, and decodes it by a CPU provided in the computer. This is realized by executing a claim detection program.
The voice reception claim detection program according to the present invention is written in a computer-readable program language, and is recorded in a computer-readable recording medium such as a magnetic disk, a CD-ROM, or a semiconductor memory. Installation on the computer is performed from these recording media or through a communication line.

  Used for telephone reception.

The function block diagram for demonstrating Example 1 of this invention. 3 is a flowchart for explaining an operation procedure according to the first embodiment. The function block diagram for demonstrating Example 2 of this invention. 9 is a flowchart for explaining an operation procedure according to the second embodiment. The function block diagram for demonstrating Example 3 of this invention. 10 is a flowchart for explaining an operation procedure of the third embodiment. The function block diagram for demonstrating Example 4 of this invention. 10 is a flowchart for explaining an operation procedure of the fourth embodiment. The function block diagram for demonstrating Example 5 of this invention. 10 is a flowchart for explaining an operation procedure of the fifth embodiment. The function block diagram for demonstrating Example 6 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Keyboard, mouse | mouth input terminal 20 Sound volume measuring part 12 Audio | voice input terminal 91 Integrated claim detection part 13 Reception start time detection part 110 Claim evaluation value normalization part 14 Competitive speech start time detection part 111 Claim detection evaluation value average memory 15 Claim detection evaluation Value measurement unit 112 Complaint detection threshold normalization unit 16 Complaint detection unit 113 Complaint detection evaluation value average update unit 17 Output unit 114 Away detection unit 19 Affinity speech recognition unit 116 Elapsed time measurement unit

Claims (18)

A reception start time detection process for detecting a reception start time from the utterance level of key input or voice input;
A mutual utterance start time detection process for detecting a mutual utterance start time;
A claim detection evaluation value measurement process for measuring a response time obtained as a difference between the summoning utterance start time and the reception start time as a claim detection evaluation value;
If the claim detection evaluation value is less than or equal to a threshold value, a claim detection process that determines that the claim is handled,
An output process for outputting operator information in response to a complaint in the claim detection process;
A method for detecting a voice reception claim, comprising: Recognition of input speech, detection of a pre-registered conflict word, and a conflict speech recognition process for obtaining a start time and an end time of the conflict word;
A claim detection evaluation value measurement process for measuring a speech utterance speed obtained by dividing the number of syllables or phonemes of the registered conflict word by the conflict word duration, as a claim detection evaluation value;
If the claim detection evaluation value is greater than or equal to a threshold value, a claim detection process that determines that the claim is handled,
An output process for outputting operator information in response to a complaint in the claim detection process;
A method for detecting a voice reception claim, comprising: Recognizing input speech, detecting pre-registered companion words, and competing utterance recognition processing to obtain the start time and end time of the companion word;
Volume measurement processing to measure the volume of the sumo utterance at the time of summon word recording,
A claim detection evaluation value measurement process for measuring a volume of the utterance utterance by dividing the volume of the utterance utterance by the duration of the affair word duration as a claim detection evaluation value;
If the claim detection evaluation value is greater than or equal to a threshold value, a claim detection process that determines that the claim is handled,
An output process for outputting operator information in response to a complaint in the claim detection process;
A method for detecting a voice reception claim, comprising: Recognizing input speech, detecting pre-registered companion words, and competing utterance recognition processing to obtain the start time and end time of the companion word;
A mutual utterance interval measurement process for measuring a mutual utterance interval obtained by subtracting the end time of the previous mutual utterance from the relevant utterance start time;
A claim detection evaluation value measurement process for measuring an average value of the interaction utterance intervals of the conflicts generated during a certain time as a claim detection evaluation value;
If the claim detection evaluation value is less than or equal to a threshold value, a claim detection process that determines that the claim is handled,
An output process for outputting the operator information in response when the claim detection unit determines that the claim is a complaint;
A method for detecting a voice reception claim, comprising:   5. A voice reception claim detection method comprising: integrated claim detection processing for detecting a claim using an integrated claim evaluation value obtained by integrating at least two or more of claim detection evaluation values according to claim 1 with a weighted linear sum. .   A result of determining that at least two of the claim detection evaluation values of claims 1 to 4 are claims is defined as an integrated claim evaluation value, and includes integrated claim detection processing for determining a claim according to the integrated claim evaluation value. A voice reception claim detection method as a feature.   Claim detection processing for performing detection by normalizing the claim detection evaluation value according to any one of claims 1 to 4 or the integrated claim evaluation value according to claim 5 or 6 and their threshold values with an average value for each input voice type A voice reception claim detection method characterized by the above.   5. A detection parameter update process for returning the average value of the claim detection evaluation value according to claim 1 or the integrated claim evaluation value according to claim 5 or 6 and the detection threshold value to a reference value after a predetermined time has elapsed. Voice reception claim detection method. A reception start time detection unit for detecting a reception start time from a key input or a speech level;
A mutual utterance start time detection unit for detecting a mutual utterance start time;
A claim detection evaluation value measuring unit that detects a response time obtained as a difference between the start time of the utterance and the reception start time as a claim detection evaluation value;
If the claim detection evaluation value is equal to or less than a threshold value, a claim detection unit that determines that the claim is handled,
An output unit that outputs operator information in response when it is determined to be a complaint in the claim detection process;
A voice reception claim detection apparatus comprising: A speech recognition unit that recognizes the input speech and detects a pre-registered conflict word and obtains a start time and an end time of the conflict word;
A claim detection evaluation value measurement unit that measures a speech utterance speed obtained by dividing the number of syllables or phonemes of the registered conflict word by the conflict word duration, as a claim detection evaluation value;
If the claim detection evaluation value is greater than or equal to a threshold, a claim detection unit that determines that the claim is handled,
An output unit that outputs operator information in response when it is determined to be a complaint in the claim detection process;
A voice reception claim detection apparatus comprising: A recognition speech recognition unit that performs input speech recognition processing, detects pre-registered conflict words, and obtains the start time and end time of the conflict words;
A volume measuring unit that measures the volume of the companion utterance when recording the companion word;
A claim detection evaluation value measuring unit that measures the volume of the utterance utterance as a claim detection evaluation value obtained by dividing the volume of the utterance utterance by the duration of the affair word; and
If the claim detection evaluation value is greater than or equal to a threshold, a claim detection unit that determines that the claim is handled,
An output unit that outputs operator information in response when the claim detection unit determines that the claim is a complaint;
A voice reception claim detection apparatus comprising: A recognition speech recognition unit that performs input speech recognition processing, detects pre-registered conflict words, and obtains the start time and end time of the conflict words;
A mutual speech interval measuring unit for measuring a mutual speech interval obtained by subtracting the end time of the previous mutual speech from the relevant speech start time;
A claim detection evaluation value measuring unit that measures an average value of the interaction speech intervals of the conflicts generated during a certain time as a claim detection evaluation value;
If the claim detection evaluation value is equal to or less than a threshold value, a claim detection unit that determines that the claim is handled,
An output unit that outputs operator information in response when the claim detection unit determines that the claim is a complaint;
A voice reception claim detection apparatus comprising:   13. A voice reception claim detection device comprising an integrated claim detection unit for detecting a claim using an integrated claim evaluation value obtained by integrating at least two or more of claim detection evaluation values according to claim 9 with a weighted linear sum. .   A result of determining that at least two of the claim detection evaluation values of claims 9 to 12 are claims is defined as an integrated claim evaluation value, and an integrated claim detection unit for determining a claim according to the integrated claim evaluation value is provided. A voice reception claim detection device as a feature.   A claim detection unit that performs detection by normalizing the claim detection evaluation value according to any one of claims 9 to 12 or the integrated claim evaluation value according to claim 5 or 6 and their threshold values with an average value for each input voice type. A voice reception complaint detection apparatus characterized by the above.   13. A voice reception claim detection device comprising: a detection parameter update unit for returning the average value of claim detection evaluation values or integrated claim evaluation values according to claim 9 to 12 and their detection threshold values to a reference value after a predetermined time has elapsed. .   A voice reception claim detection program, which is written in a computer-readable program language, and causes the computer to execute the voice reception claim detection method according to any one of claims 1 to 8.   A recording medium comprising a computer-readable recording medium, wherein the voice reception claim detection program according to claim 17 is recorded on the recording medium.

Images