JP2010021681A - Reception flow creation program, reception flow creation method, and reception flow creating apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a reception flow creation program, a reception flow creation method, and a reception flow creating apparatus which can appropriately treat silent interval which exists during a speech, and which easily extract problems in each procedures. <P>SOLUTION: The problem can be solved by a computer which is made to function as: a procedure keyword table recording means; an utterance interval extraction means which extracts the utterance interval of a customer and a reception person; a voice recognition means which carries out voice recognition of the speech recording data; a procedure extraction means which extracts a procedure, corresponding to a recognized keyword included in the voice recognition result based on the procedure keyword table; a procedure time calculation means which calculates the time from a start, until an end of the procedure, and at the same time, if there is a silent interval other than the speech interval, makes a talker, immediately after the silent interval as a generator of the silent interval, and adds the time of the silent interval to the time from the start, until the end of the procedure; and a reception flow creation means which creates and outputs the reception flow, based on the time from the start, until the end of the procedure. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a reception flow creation program, a reception flow creation method, and a reception flow creation device, and more particularly to a reception flow creation program, a reception flow creation method, and a reception flow for creating a reception flow from call recording data obtained by recording a call to a customer. The present invention relates to a flow creation device.

  For example, in a call center that answers a question from a customer such as a help desk, the answering time for the same question varies depending on an agent (agent). This is thought to be because the procedure for separating the situation / factor of the question contents for each agent and obtaining the essential problem of the question and the procedure for explaining the answer are different.

  On the other hand, the call center manager wants to shorten the time required for one response without impairing the quality of the response and to improve the efficiency of the call center operations. For this reason, the administrator wants to identify the situation / factors of the question content, clarify the difference between the procedure for obtaining the essential problem of the question and the procedure for explaining the answer, and make it possible for any agent to answer in the optimal procedure.

  At the current call center, the customer's question (for example, the flexible disk cannot be read) and the final answer to the question (for example, xxx.EXE of the program disk is executed, and the flexible disk can be opened after rebooting. I have confirmed this), and is recorded as a response record.

  However, there is no history in the response record until the final answer. For example, xxx. Although it is described that EXE has been executed, the procedure of xxx. It is not described whether EXE is executed. In this way, from the remaining response records, we did not know the procedure (question isolation procedure) to separate the situation / factors of the question contents and obtain the essential question of the question, and the procedure (answer procedure) to explain the answer. .

  By the way, in the call center, all calls corresponding to each customer are recorded and left as call recording data. By listening to the call recording data, the call center can understand the problem isolation procedure and the answer procedure (explanation procedure) later. However, it takes an enormous amount of time for humans to actually listen to call recording data and examine the problem isolation procedure and the explanation procedure.

As a technique for extracting a characteristic part from voice dialogue data such as call recording data, there is, for example, Patent Document 1, and there is a technique for extracting a question-response part using the utterance characteristic. Further, as a technique using call information in a call center, for example, there is Patent Document 2, which measures the processing time for one case from the time of a call related thereto.
JP 2005-242891 A JP 2003-298748 A

  In order to extract each procedure such as the explanation procedure from the call recording data and extract the time required for each procedure, means for extracting the procedure from the call recording data, means for calculating the time required for each procedure, A means for processing a silent section (call interruption section) peculiar to a call is necessary. In particular, in a conventional call center, it has not been studied how to process a silent section peculiar to a call.

  During a call between an agent and a customer in a call center, there is a silent section where the call is interrupted. The reason for the silent section is that the agent is waiting for the customer because the agent is examining the response, the customer is waiting for the customer because the agent is searching for customer information (such as past response contents), and the customer is asking The agent is waiting to check the details of the content, the customer is waiting for the agent because the customer is actually doing the action explained to the agent, etc. It is different each time.

  Therefore, when there is a silent section during a call, it is necessary to determine which procedure is the silent section caused by which procedure, which time is included in the required time, and who the silent section is caused by .

  One embodiment of the present invention has been made in view of the above points, and it is possible to appropriately handle a silent section existing during a call and to easily extract a problem of each procedure. An object is to provide a reception flow creation method and a reception flow creation apparatus.

  In order to solve the above-described problem, an embodiment of the present invention describes a computer for creating a reception flow from call recording data obtained by recording a reception call for a customer, and the procedure for each reception procedure for a customer. Therefore, a procedure keyword table recording means for pre-recording a keyword as a procedure keyword table in association with a keyword spoken by a responder and a keyword spoken by a customer in response to the description of the procedure, and extracted from the call recording data Speech segment extraction means for extracting speech segments of customers and respondents based on prosodic information, and speech recognition means for speech recognition of the call recording data and outputting the recognized keyword, speaker, and appearance time as speech recognition results And reading the procedure keyword table, the procedure corresponding to the recognized keyword included in the speech recognition result, Corresponding to the procedure extraction means for extracting based on the forward keyword table, the utterance section extracted by the utterance section extraction means, the speech recognition result output by the speech recognition means, and the recognized keyword extracted by the procedure extraction means Based on the procedure, the time from the start to the end of the procedure is calculated, and when there is a silent section other than the speech section, the speaker immediately after the silent section is set as the generator of the silent section, and the silence A procedure time calculating means for adding the time of the section to a time from the start to the end of the procedure, and a response flow creating means for creating and outputting the response flow based on the time from the start to the end of the procedure This is a response flow creation program.

  In addition, what applied the component, the expression, or arbitrary combinations of the component of one Embodiment of this invention to a method, an apparatus, a system, a computer program, a recording medium, a data structure, etc. is also effective as an aspect of this invention. .

  As described above, according to one embodiment of the present invention, a reception flow creation program and a reception flow creation that can appropriately handle a silent section existing during a call and can easily extract the problems of each procedure. A method and a service flow creation apparatus can be provided.

  Next, the best mode for carrying out the present invention will be described based on the following embodiments with reference to the drawings.

  FIG. 1 is an explanatory diagram showing an example of an outline of the present embodiment. The response flow creation apparatus 1 of the present embodiment receives call recording data obtained by recording all calls to each customer. The response flow creation device 1 specifies a procedure location from the input call recording data as described later, and calculates the time required for each procedure.

  When there is a silent section before and after the specified procedure location, the reception flow creation apparatus 1 determines which procedure is the silent section caused by the procedure, and in which procedure the silent section is put as described below. In addition, the reception flow creation device 1 creates a reception flow (a flow of each procedure such as an explanation procedure) from the identified procedure location and the time required for each procedure as described later. The corresponding flow creation apparatus 1 outputs the flow of each procedure and the time required for each procedure.

  FIG. 2 is an image diagram showing an example of processing for specifying a procedure part from input call recording data and calculating a time required for each procedure. When the response flow creation device 1 specifies a procedure location from the input call recording data, the call recording is performed at a location where a keyword (KW) indicating an agent procedure and a customer keyword corresponding to the agent keyword appear. Search from data. The response flow creation device 1 includes the time (required time) required for the procedure indicated by the keyword from the start of the utterance section including the agent keyword to the end of the utterance section including the customer keyword corresponding to the agent keyword. To do.

  For example, in FIG. 2, from the start of the utterance section 10 including the agent keyword “program disk, double click”, the end of the utterance section 11 including the customer keyword “program disk” corresponding to the agent keyword “program disk, double click”. Is the time required for the procedure indicated by the keyword “program disk, double click”.

  FIG. 3 is an image diagram illustrating an example of a process when there is a gap among processes for calculating the time required for each procedure. Since the response flow creation device 1 cannot reliably extract all keywords by voice recognition, if there is a gap in the identified procedure, the procedure in the gap is estimated.

  For example, in FIG. 3, from the start of the utterance section 20 including the agent keyword “program disk, double click” to the end of the utterance section 21 including the customer keyword “yes” corresponding to the agent keyword “program disk, double click”. Is the procedure 2 indicated by the keyword “program disk, double click”.

  Further, in FIG. 3, from the start of the utterance section 24 including the agent keyword “folder, open”, the end of the utterance section 25 including the customer keyword “opened” corresponding to the agent keyword “folder, open”. Step 4 is the procedure 4 indicated by the keyword “open folder”.

  If it is known that there is a procedure 3 between the procedure 2 and the procedure 4, the response flow creation device 1 starts from the start of the utterance section 22 in the gap between the procedure 2 and the procedure 4 until the end of the utterance section 23. Is estimated as procedure 3.

  FIG. 4 is an image diagram showing an example of a process for a silent section among the processes for calculating the time required for each procedure. The response flow creation device 1 identifies which procedure corresponds to a silent section or a valid silent section, depending on where the silent section appears.

  For example, in FIG. 4, from the start of the utterance section 30 including the agent keyword “program disk, double click” to the end of the utterance section 32 including the customer keyword “yes” corresponding to the agent keyword “program disk, double click”. Is the procedure indicated by the keyword “program disk, double click”.

  The procedure indicated by the keyword “program disc, double click” includes a silent section 33 between the utterance section 31 and the utterance section 32. The response flow creation device 1 identifies who caused the silent section 33 by who the first speaker immediately after the silent section 33 is.

  Specifically, the response flow creation apparatus 1 estimates that the silent section 33 is caused by the customer because the speaker in the speech section 32 immediately after the silent section 33 is a customer. When the speaker in the utterance section 32 immediately after the silence section 33 is an agent, the response flow creation apparatus 1 estimates that the silence section 33 is caused by the agent.

  FIG. 5 is a conceptual diagram showing another example of the process for the silent section among the processes for calculating the time required for each procedure. In FIG. 5, a silent section 44 appears after the utterance section 41 including the customer keyword “yes” corresponding to the agent keyword “program disc, double click”.

  The response flow creation device 1 determines whether the silent section 44 is appropriate, as described later, depending on the presence or absence of an agent operation accompanying the procedure indicated by the agent keyword “program disk, double click”. It is determined whether it is considered as a part of the procedure indicated by the keyword “program disk, double click”.

  Specifically, if there is a post-operation of the agent operation accompanying the procedure from the start of the utterance section 40 to the end of the utterance section 41, the response flow creation apparatus 1 estimates that the work is performed in the silent section 44, The silent section 44 is added to the time required for the procedure from the start of the utterance section 40 to the end of the utterance section 41. Further, the response flow creation device 1 determines that the silent section 44 is appropriate if there is a post-operation of the agent operation accompanying the procedure from the start of the utterance section 40 to the end of the utterance section 41.

  Note that the response flow creation apparatus 1 has no post-operation of the agent operation associated with the procedure from the start of the utterance section 40 to the end of the utterance section 41, and is associated with the procedure from the start of the utterance section 42 to the end of the utterance section 43. If there is no work prior to the agent operation, the silent section 44 is added to the time required for the procedure from the start of the utterance section 40 to the end of the utterance section 41. Moreover, the reception flow creation apparatus 1 determines that the silent section 44 is not valid.

  FIG. 6 is a hardware configuration diagram of an example of a reception flow creation apparatus. The response flow creation device 1 includes an input device 51, an output device 52, a drive device 53, an auxiliary storage device 54, a main storage device 55, an arithmetic processing device 56, and an interface device 57 that are mutually connected by a bus B. Configured.

  The input device 51 includes a keyboard and a mouse, and is used for inputting various signals. The output device 52 includes a display device and is used to display various windows, data, and the like. The interface device 57 includes a modem, a LAN card, and the like, and is used for connecting to a network.

  The reception flow creation program of this embodiment is at least a part of various programs that control the reception flow creation apparatus 1. The response flow creation program is provided by, for example, distribution of the recording medium 58 or downloading from a network. The recording medium 58 on which the response flow creation program is recorded is a recording medium such as a CD-ROM, a flexible disk, a magneto-optical disk, etc. for recording information optically, electrically or magnetically, a ROM, a flash memory, etc. Various types of recording media such as a semiconductor memory for electrically recording information can be used.

  When the recording medium 58 on which the response flow creation program is recorded is set in the drive device 53, the response flow creation program is installed from the recording medium 58 to the auxiliary storage device 54 via the drive device 53. The response flow creation program downloaded from the network is installed in the auxiliary storage device 54 via the interface device 57. The auxiliary storage device 54 stores the installed response flow creation program and stores necessary files, data, and the like.

  The main storage device 55 reads the response flow creation program from the auxiliary storage device 54 and stores it when the response flow creation program is started. The arithmetic processing unit 56 implements various processes as described later by a processing unit (specific means) in which software and hardware resources cooperate in accordance with a reception flow creation program stored in the main storage device 55. Yes.

  FIG. 7 is a configuration diagram showing the relationship between the processing unit and data of the reception flow creation apparatus. 7 includes a prosodic information extraction unit 61, a speech segment extraction unit 62, a phase estimation unit 63, a speech recognition unit 64, a procedure extraction unit 65, a procedure time calculation unit 66, and a response. The flow creation unit 67 is included.

  Also, the response flow creation device 1 of FIG. 7 includes call recording data 71, prosody data 72, speech segment information 73, phase information 74, speech recognition keyword list 75, speech recognition result 76, procedure keyword table 77, procedure extraction data 78. , Reception flow data 79 and reception procedure time information 80 are used.

  Here, the relationship between the processing unit of the reception flow creation apparatus 1 and data will be described with reference to the flowchart of FIG. FIG. 8 is a flowchart of an example showing the processing procedure of the reception flow creation apparatus.

  Proceeding to step S 1, the prosody information extracting unit 61 receives the call recording data 71. The call recording data 71 is, for example, a stereo file in which agent voice and customer voice are recorded on different channels. In the call recording data 71, all calls from the beginning to the end of each call are recorded. For example, the call recording data 71 is accompanied by information such as an incident ID and recording start date / time associated with the response recording.

  The prosody information extraction unit 61 extracts agent and customer prosody data 72 from the input call recording data 71 and outputs it. Details of the processing of the prosody information extraction unit 61 and the prosody data 72 will be described later.

  In step S 2, the utterance section extraction unit 62 extracts the utterance sections of the agent and the customer from the input prosodic data 72 and outputs the extracted utterance section information 73. Details of the processing of the utterance section extraction unit 62 and the utterance section information 73 will be described later. Further, the phase estimation unit 63 proceeds to step S 3, estimates the question phase and the problem isolation / answer phase from the input utterance section information 73, and outputs it as phase information 74. Note that the processing by the phase estimation unit 63 is not essential. Details of the processing of the phase estimation unit 63 and the phase information 74 will be described later.

  In step S4, the voice recognition unit 64 receives the call recording data 71, the phase information 74, and the voice recognition keyword list 75. The voice recognition unit 64 performs voice recognition of the call recording data 71 using the input phase information 74 and the voice recognition keyword list 75 and outputs a voice recognition result 76. Details of the processing of the voice recognition unit 76, the voice recognition keyword list 75, and the voice recognition result 76 will be described later.

  In step S5, the procedure extraction unit 65 receives the speech recognition result 76 and the procedure keyword table 77. The procedure extraction unit 65 specifies the procedure part using the input speech recognition result 76 and the procedure keyword table 77 and outputs the procedure extraction data 78. Details of the process of the procedure extraction unit 65, the procedure keyword table 77, and the procedure extraction data 78 will be described later.

  In step S6, the procedure time calculation unit 66 receives the utterance section information 73, the speech recognition result 76, and the procedure keyword table 77. The procedure time calculation unit 66 calculates the procedure time using the input utterance section information 73, the speech recognition result 76, and the procedure keyword table 77, and adds data to the procedure extraction data 78. Details of the process of the procedure time calculation unit 66 and the data added to the procedure extraction data 78 will be described later.

  In step S 7, the response flow creation unit 67 receives the procedure extraction data 78. The reception flow creation unit 67 creates and outputs reception flow data 79 and reception procedure time information 80 using the input procedure extraction data 78. Details of the processing of the reception flow creation unit 67, reception flow data 79, and reception procedure time information 80 will be described later.

  Below, the process which each process part of the reception flow preparation apparatus 1 shown in FIG. 7 performs is demonstrated in detail.

(Prosodic information extraction unit 61)
The prosodic information extraction unit 61 receives call voice data 71 which is stereo data obtained by recording agent voice and customer voice on different channels. The prosody information extraction unit 61 calculates the prosody information (power value) of the input call voice data 71 at regular intervals, and extracts and outputs the prosody data 72 of the agent and the customer.

  FIG. 9 is an explanatory diagram showing an example of prosodic data of agents and customers. In the prosody data 72 shown in FIG. 9, one row is a set of time and power value, and represents a power value string for every fixed time (12.8 msec).

(Speech section extraction unit 62)
The utterance section extraction unit 62 receives the prosodic data 72 of the agent and customer. The utterance section extraction unit 62 extracts a point where the power value equal to or greater than the threshold is continuous from the input prosodic data 72 of the agent and the continuous time is equal to or greater than the minimum utterance time as the agent utterance section. Output as utterance section information 73.

  Further, the utterance section extraction unit 62 extracts a portion where the power value equal to or higher than the threshold is continuous from the input customer prosodic data 72 and the continuous time is equal to or longer than the minimum utterance time as the customer's utterance section. Is output as utterance section information 73.

  FIG. 10 is an explanatory diagram showing an example of agent and customer utterance section information. The utterance section information 73 shown in FIG. 10 is a set of the utterance ID, start time, and end time, and represents the start time and end time for each utterance section.

(Phase estimation unit 63)
The processing performed by the phase estimation unit 63 is not essential. The processing performed by the phase estimation unit 63 contributes to accuracy improvement. The phase estimation unit 63 receives the utterance section information 73 of the agent and the customer. The phase estimation unit 63 estimates the initiative speaker of the utterance section from the input agent and customer utterance section information 73. In addition, although the method of estimating the initiative speaker of an utterance area is known, the method described in Unexamined-Japanese-Patent No. 2007-184699 can be utilized, for example.

  The phase estimation unit 63 estimates the utterance section in which the customer immediately after the start of the call recording data 71 is the initiative speaker as the question phase. Further, the phase estimation unit 63 estimates the utterance section in which the agent after the question phase is the initiative speaker as the problem isolation / answer phase. The phase estimation unit 63 outputs the estimated question phase and the problem isolation / answer phase as phase information 74.

  FIG. 11 is an explanatory diagram showing an example of phase information. The phase information 74 shown in FIG. 11 is a set of phase, start time and end time. The phase information 74 represents a start time and an end time for each question phase and question isolation / answer phase.

(Voice recognition unit 64)
The voice recognition unit 64 receives the call recording data 71 and the voice recognition keyword list 75. When the phase information 74 is used, the speech recognition unit 64 also receives the phase information 74.

  FIG. 12 is an explanatory diagram showing an example of a voice recognition keyword list. The voice recognition keyword list 75 is a voice recognition dictionary. The speech recognition keyword list 75 is created as described later from a manual, FAQ, procedure manual, and the like. The speech recognition keyword list 75 is prepared for each agent and customer. The voice recognition keyword list 75 shown in FIG. 12 is a set of keywords and readings.

  The voice recognition unit 64 uses word spotting for voice recognition. The voice recognition unit 64 performs voice recognition of the call recording data 71. Note that when the phase information 74 is used, voice recognition is performed only in the separation / answer phase of the call recording data 71. For example, in the case of call voice data 71 in which the voice of an agent is recorded, the voice recognition unit 64 uses the agent voice recognition keyword list 75. The voice recognition unit 64 uses the customer voice recognition keyword list 75 in the case of the call voice data 71 in which the customer voice is recorded.

  The voice recognition unit 64 outputs a voice recognition result 76 that is a result of voice recognition of the call recording data 71. FIG. 13 is an explanatory diagram showing an example of a speech recognition result. The speech recognition result 76 shown in FIG. 13 is a set of a recognition result ID, a recognition keyword, a speaker, and an appearance time. The voice recognition result 76 represents a recognition keyword, a speaker, and an appearance time for each recognition result ID.

  The recognition result ID is for identifying the voice recognition result. The recognition keyword represents a recognition keyword recognized by voice recognition. The speaker represents an agent or a customer who has spoken the recognition keyword. The appearance time represents the elapsed time from the beginning of the call recording data 71 in which the recognition keyword appears.

(Procedure extraction unit 65)
The procedure extraction unit 65 receives the speech recognition result 76 and the procedure keyword table 77. FIG. 14 is an explanatory diagram showing an example of a procedure keyword table. The procedure keyword table 77 in FIG. 14 is a set of procedure ID, agent operation, extracted keyword, part of speech, and corresponding phrase. The procedure ID identifies the procedure. The agent operation represents the operation content of the agent in each procedure.

  The extracted keyword is a keyword for specifying the procedure that the agent speaks to explain the procedure. The part of speech is the part of speech information of the extracted keyword. The corresponding word / phrase is a word / phrase uttered by a customer response to the agent's procedure explanation. The procedure keyword table 77 is created from a manual, FAQ, procedure manual, etc. as described later.

  The procedure extraction unit 65 identifies the procedure part as follows. In the procedure extraction unit 65, the speaker sequentially extracts the recognition keyword of the agent as the extraction keyword from the input speech recognition result 76. The procedure extracting unit 65 extracts the procedure ID and the corresponding phrase corresponding to the extracted keyword of the extracted agent from the procedure keyword table 77.

  The procedure extraction unit 65 extracts from the voice recognition result 76 the speaker immediately after the extracted keyword of the agent as a customer keyword that is a customer's recognized keyword and is a corresponding phrase. The term “immediately after the agent extraction keyword” here refers to the time period within a predetermined time until the next agent recognition keyword appears.

  The procedure extraction unit 65 sets the extracted keyword of the agent, the customer keyword in which the speaker immediately after the extracted keyword of the agent is the customer's recognized keyword and the corresponding phrase, the procedure ID corresponding to the extracted keyword, and the recognition result ID as one. The procedure part is specified as an indication and written in the procedure extraction data 78.

  FIG. 15 is an explanatory diagram showing an example of procedure extraction data entered by the procedure extraction unit. The procedure extraction data 78 of FIG. 15 includes an agent keyword, an agent keyword recognition result ID, a customer keyword, a customer keyword recognition result ID, a procedure ID, a start time, an end time, a silence interval, a silence interval generator, and a silence interval. Is a pair of validity.

  The procedure extraction data 78 in FIG. 15 represents a state in which an agent keyword, an agent keyword recognition result ID, a customer keyword, a customer keyword recognition result ID, and a procedure ID are entered.

  FIG. 16 is a flowchart illustrating an example of the processing procedure of the procedure extracting unit. In step S11, the procedure extracting unit 65 sequentially extracts the recognition keyword of the agent from the speech recognition result 76 as the extracted keyword. The procedure extracting unit 65 extracts the procedure ID and the corresponding phrase corresponding to the extracted keyword of the extracted agent from the procedure keyword table 77.

  In step S12, the procedure extraction unit 65 attempts to acquire the next recognition keyword whose speaker is a customer from the speech recognition result 76. Proceeding to step S13, if there is a next recognition keyword whose speaker is a customer from the speech recognition result 76, the procedure extraction unit 65 acquires the next recognition keyword whose speaker is a customer, and proceeds to the processing of step S14. .

  In step S14, the procedure extraction unit 65 determines that the appearance time of the acquired next recognition keyword of the customer is within the range of the appearance time of the recognition keyword of the agent + α (predetermined time) and the next recognition of the agent. It is determined whether it is before the appearance time of the keyword and the next recognition keyword of the acquired customer is a corresponding phrase of the recognition keyword of the agent.

  The appearance time of the next recognition keyword of the acquired customer is within the range of the appearance time of the recognition keyword of the agent + α (predetermined time), and is earlier than the appearance time of the next recognition keyword of the agent. If the next recognized keyword of the customer is the corresponding phrase of the recognized keyword of the agent, the procedure extracting unit 65 proceeds to step S15.

  The appearance time of the next recognition keyword of the acquired customer is within the range of the appearance time of the recognition keyword of the agent + α (predetermined time), and is earlier than the appearance time of the next recognition keyword of the agent. If the next recognized keyword of the customer is not the corresponding phrase of the recognized keyword of the agent, the procedure extracting unit 65 returns to step S12.

  In step S15, the procedure extraction unit 65 sets the agent keyword, the recognition result ID of the agent keyword, the customer keyword, the recognition result ID of the customer keyword, and the procedure ID as one procedure part. Fill in 78. The processes of steps S11 to S15 are repeated as long as there is an unprocessed speech recognition result 76.

(Procedure time calculation unit 66)
The procedure time calculation unit 66 receives the utterance section information 73, the speech recognition result 76, and the procedure keyword table 77. The procedure time calculation unit 66 calculates the time for each procedure according to the processing procedure shown in FIG.

  FIG. 17 is a flowchart illustrating an example of a processing procedure performed by the procedure time calculation unit. In step S21, the procedure time calculation unit 66 obtains the start time and end time of each procedure entered in the procedure extraction data 78 by the procedure extraction unit 65. Details of the process in step S21 will be described later.

  Proceeding to step S22, the procedure time calculation unit 66 summarizes the same procedure that is consecutive among the procedures entered in the procedure extraction data 78 as a duplicate procedure, and estimates and inserts the procedure of extraction omission. . Details of the process in step S22 will be described later. Proceeding to step S23, the procedure time calculation unit 66 performs a silent section. Details of step S23 will be described later.

  The process of step S21 is performed as follows. The process of step S21 is to enter the start time and end time for all the procedures entered in the procedure extraction data 78 by the procedure extraction unit 65.

  First, the procedure time calculation unit 66 sequentially extracts all procedures entered in the procedure extraction data 78 by the procedure extraction unit 65. The procedure time calculation unit 66 acquires the appearance time of the agent keyword from the speech recognition result 76 based on the recognition result ID of the agent keyword of the extracted procedure. Then, the procedure time calculation unit 66 acquires an utterance section including the appearance time of the agent keyword from the utterance section information 73 of the agent, and uses the start time of the utterance section as the procedure start time.

  The procedure time calculation unit 66 acquires the appearance time of the customer keyword from the speech recognition result 76 based on the recognition result ID of the customer keyword of the extracted procedure. Then, the procedure time calculation unit 66 acquires the utterance section including the appearance time of the customer keyword from the customer utterance section information 73, and sets the end time of the utterance section as the end time of the procedure.

  Then, the procedure time calculation unit 66 enters the procedure start time and end time in the procedure extraction data 67. FIG. 18 is an explanatory diagram showing an example of procedure extraction data in which the start time and end time are entered by the procedure time calculation unit.

  FIG. 19 is a flowchart illustrating an example of the processing procedure of step S21. In step S31, the procedure time calculation unit 66 acquires the appearance time of the agent keyword from the speech recognition result 76 based on the recognition result ID of the agent keyword of the procedure. The procedure time calculation unit 66 acquires an utterance section including the appearance time of the agent keyword from the utterance section information 73 of the agent, and uses the start time of the utterance section as the procedure start time.

  In step S32, the procedure time calculation unit 66 acquires the appearance time of the customer keyword from the voice recognition result 76 based on the recognition result ID of the customer keyword of the procedure. Then, the procedure time calculation unit 66 acquires the utterance section including the appearance time of the customer keyword from the customer utterance section information 73, and sets the end time of the utterance section as the end time of the procedure.

  In step S 33, the procedure time calculation unit 66 writes the procedure start time and end time in the procedure extraction data 67. In step S 34, the procedure time calculation unit 66 advances the procedure to be processed to the procedure following the procedure extraction data 78. The processes of steps S31 to S34 are repeated as long as there are unprocessed procedures in the procedure extraction data 78.

  The process of step S22 is performed as follows. The process of step S22 is to insert duplicate procedure corrections and extraction omission procedures for all procedures entered in the procedure extraction data 78 by the procedure extraction unit 65.

  First, the procedure time calculation unit 66 sequentially extracts all procedures entered in the procedure extraction data 78 by the procedure extraction unit 65. The procedure time calculation unit 66 checks whether the procedure ID of the procedure to be processed is the same as the procedure ID of the next procedure. When the procedure ID of the procedure to be processed and the procedure ID of the next procedure are the same, the procedure time calculation unit 66 merges the procedure to be processed and the next procedure. At this time, the procedure time calculation unit 66 adopts the start time of the procedure to be processed as the start time of the procedure, and adopts the end time of the next procedure as the end time of the procedure.

  The procedure time calculation unit 66 merges the agent keyword, the agent keyword recognition result ID, the customer keyword, and the customer keyword recognition result ID with the procedure to be processed and the next procedure, and deletes the merged next procedure. .

  Further, the procedure time calculation unit 66 has a procedure between the procedure ID of the procedure to be processed and the procedure ID of the next procedure, and between the end of the procedure to be processed and the start of the next procedure. Check if there is an utterance interval. It can be checked from the procedure keyword table 77 whether or not a procedure exists between the procedure ID of the procedure to be processed and the procedure ID of the next procedure. Whether or not there is an utterance interval between the end of the procedure to be processed and the start of the next procedure can be checked from the utterance interval information 73.

  If there is a procedure between the procedure ID of the procedure to be processed and the procedure ID of the next procedure, and there is an utterance section between the end of the procedure to be processed and the start of the next procedure, the procedure time The calculation unit 66 estimates that there is an extraction failure procedure, inserts the extraction failure procedure between the procedure to be processed in the procedure extraction data 78 and the next procedure, and ends the processing target procedure and the next procedure. The start time and end time of the utterance section between the start and the start time and end time of the procedure are entered.

  FIG. 20 is an explanatory diagram illustrating an example of procedure extraction data in which the procedure time calculation unit corrects duplicate procedures and inserts an extraction failure procedure. In the procedure extraction data 78 of FIG. 20, the procedures with the procedure ID “A-1” are collected as the same procedure, and the procedure with the procedure ID “A-3” is inserted as a procedure for omission of extraction.

  FIG. 21 is a flowchart illustrating an example of the processing procedure of step S22. In step S41, the procedure time calculation unit 66 determines whether or not the procedure to be processed is the same as the next procedure.

  If the procedure to be processed and the next procedure are the same, the procedure time calculation unit 66 proceeds to step S42 and merges the procedure to be processed and the next procedure. At this time, the procedure time calculation unit 66 adopts the start time of the procedure to be processed as the start time of the procedure, adopts the end time of the next procedure as the end time of the procedure, and proceeds to step S43. If the procedure to be processed is not the same as the next procedure, the procedure time calculation unit 66 proceeds from step S41 to step S43.

  In step S43, the procedure time calculation unit 66 determines whether there is a procedure between the procedure to be processed and the next procedure, and whether there is an utterance section between the procedure to be processed and the next procedure. Determine whether.

  If there is a procedure between the procedure to be processed and the next procedure, and there is an utterance section between the procedure to be processed and the next procedure, the procedure time calculation unit 66 proceeds to step S44 and extracts Estimated that there is a leaking procedure, insert one line between the procedure to be processed in the procedure extraction data 78 and the next procedure, and enter the procedure ID of the procedure existing between the procedure to be processed and the next procedure Then, the start time and end time of the utterance section between the procedure to be processed and the next procedure are entered as the start time and end time of the procedure, and the process proceeds to step S45.

  When there is no procedure between the procedure to be processed and the next procedure, or when there is no utterance section between the procedure to be processed and the next procedure, the procedure time calculation unit 66 starts from step S43. Proceed to step S45. In step S 45, the procedure time calculation unit 66 advances the procedure to be processed to the procedure following the procedure extraction data 78. The processes of steps S41 to S45 are repeated as long as there are unprocessed procedures in the procedure extraction data 78.

  The process of step S23 is performed as follows. The process of step S23 performs a silent section. First, the procedure time calculation unit 66 creates silent section information. The procedure time calculation unit 66 obtains each speaker's utterance section in turn from the utterance section information 73, and if a section in which no utterance exists is a certain time or more, a section in which there is no utterance for a certain period or more is a silent section. Silent section information is created. The fixed time for determining the silent section is a time that is allowed as a time naturally occurring in a normal conversation.

  FIG. 22 is an explanatory diagram showing an example of silent section information. The silent section information shown in FIG. 22 is a set of start time, end time and section, and represents the start time, end time and section for each silent section.

  Next, the procedure time calculation unit 66 determines whether there is a silent section in the procedure to be processed by looking at the start time and end time of the procedure to be processed. If there is a silent section, the procedure time calculation unit 66 determines the speaker that generated the silent section (the generator of the silent section) from the speaker immediately after the appearance of the silent section and the accompanying work (agent operation). It is determined whether or not the silent section may occur (a procedure with validity).

  The procedure time calculation unit 66 fills in the silence section of the procedure extraction data 78, the generator of the silence section, and the validity of the silence section from the determined result.

  Further, the procedure time calculation unit 66 determines whether there is a silent section between the procedure to be processed and the next procedure by looking at the start time of the procedure to be processed and the end time of the next procedure. If there is a silent section, the procedure time calculation unit 66 generates a silent section from the speaker immediately after the appearance of the silent section and the work (agent operation) associated with the target procedure and the next procedure. , Start time and end time to be corrected accordingly, the speaker that caused the silent period, and the validity of the silent period.

  The procedure time calculation unit 66 fills in the silence section of the procedure extraction data 78, the generation of the silence section, and the validity of the silence section from the determined result, and corrects the start time and end time if necessary.

  FIG. 23 is an explanatory diagram showing an example of procedure extraction data obtained by processing a silent section. In the procedure extraction data 78 of FIG. 23, a silence ID, a generator of the silence interval, and the validity of the silence interval are entered in the procedure having the procedure ID “A-2”.

  FIG. 24 is a flowchart illustrating an example of the processing procedure of step S23. In step S51, the procedure time calculation unit 66 creates silent section information. In step S52, the procedure time calculation unit 66 determines whether there is a silent section from the start time to the end time of the procedure to be processed. If there is a silent section, the procedure time calculation unit 66 proceeds to step S53, performs a silent section process A described later, and then proceeds to step S54.

  Proceeding to step S54, the procedure time calculation unit 66 fills in the silence section of the procedure extraction data 78, the occurrence of the silence section, and the validity of the silence section from the result of the process A of the silence section of step S53, and then proceeds to step S56. move on. If there is no silence section from the start time to the end time of the procedure to be processed, the procedure time calculation unit 66 proceeds to step S55, and the silence period of the procedure extraction data 78 indicates that the time of the silence section is zero. After entering “0”, the process proceeds to step S56.

  In step S56, the procedure time calculation unit 66 determines whether there is a silent section between the start time of the procedure to be processed and the end time of the procedure. If there is a silent section, the procedure time calculation unit 66 proceeds to step S57, performs a silent section processing B described later, and then proceeds to step S58. Proceeding to step S58, the procedure time calculation unit 66 fills in the silence interval of the procedure extraction data 78, the occurrence of the silence interval, and the validity of the silence interval from the result of the silence interval process B of step S57, and is necessary. If the start time and end time are corrected, the process proceeds to step S60.

  If there is no silence interval between the start time of the procedure to be processed and the end time of the next procedure, the procedure time calculation unit 66 proceeds to step S59 and the time of the silence interval is included in the silence interval of the procedure extraction data 78. After “0” representing 0 is entered, the process proceeds to step S60. In step S 60, the procedure time calculation unit 66 advances the procedure to be processed to the procedure following the procedure extraction data 78. The processes of steps S52 to S60 are repeated as long as there are unprocessed procedures in the procedure extraction data 78.

  FIG. 25 is a flowchart illustrating an example of the processing procedure of step S53. In step S71, the procedure time calculation unit 66 determines whether or not the speaker immediately after the appearance of the silent section is an agent. If the speaker immediately after the appearance of the silent section is an agent, the procedure time calculation unit 66 sets the speaker that generated the silent section as an agent, and proceeds to step S73.

  In step S73, the procedure time calculation unit 66 refers to the procedure keyword table 77 and determines whether the agent operation of the procedure to be processed is “None” indicating empty. If the agent operation of the procedure to be processed is empty, the procedure time calculation unit 66 proceeds to step S74 and determines that the silence section is not valid.

  In step S71, if the speaker immediately after the appearance of the silent section is a customer, the procedure time calculation unit 66 sets the speaker who generated the silent section as the customer. In step S73, if the agent operation of the procedure to be processed is not empty, the procedure time calculation unit 66 proceeds to step S76.

  In step S76, the procedure time calculation unit 66 refers to the procedure keyword table 77 and determines whether the agent operation of the procedure to be processed is other than “intermediate processing”. If the agent operation of the procedure to be processed is other than “intermediate processing”, the procedure time calculation unit 66 proceeds to step S74 and determines that the silence section is not valid. If the agent operation of the procedure to be processed is “intermediate processing”, the procedure time calculation unit 66 proceeds to step S77 and determines that the silence section is valid.

  FIG. 26 is a flowchart showing an example of the processing procedure of step S57. In step S81, the procedure time calculation unit 66 determines whether or not the speaker immediately after the appearance of the silent section is an agent. If the speaker immediately after the appearance of the silent section is an agent, the procedure time calculation unit 66 sets the speaker that generated the silent section as an agent, and proceeds to step S83.

  Proceeding to step S83, the procedure time calculation unit 66 refers to the procedure keyword table 77 and determines whether the agent operation of the procedure to be processed is “post-processing” representing empty. If the agent operation of the procedure to be processed is “post-processing”, the procedure time calculation unit 66 proceeds to step S84, corrects the end of the procedure to be processed to the end time of the silence interval, and targets the silence interval as a target. Add to the procedure. Proceeding to step S85, the procedure time calculation unit 66 determines that the silence section is valid.

  In step S81, if the speaker immediately after the appearance of the silent section is a customer, the procedure time calculation unit 66 sets the speaker who generated the silent section as the customer. If the agent operation of the procedure to be processed is not “post-processing” in step S83, the procedure time calculation unit 66 proceeds to step S87.

  In step S87, the procedure time calculation unit 66 refers to the procedure keyword table 77 and determines whether the agent operation of the procedure to be processed is “pre-processing”.

  If the agent operation of the procedure to be processed is “pre-processing”, the procedure time calculation unit 66 proceeds to step S88, corrects the start of the procedure to be processed to the start time of the silence interval, and sets the silence interval to the next time. Add to the procedure. Proceeding to step S85, the procedure time calculation unit 66 determines that the silence section is valid.

  If the agent operation of the procedure to be processed is not “pre-processing”, the procedure time calculation unit 66 proceeds to step S89, corrects the end of the procedure to be processed to the end time of the silence interval, and sets the silence interval to the next time. Add to the procedure. Proceeding to step S90, the procedure time calculation unit 66 determines that the silence section is not valid.

(Responding flow creation unit 67)
The procedure extraction data 78 is input to the reception flow creation unit 67. The reception flow creation unit 67 creates the reception flow data 79 of FIG. 27 in order from the procedure extraction data 78 with one procedure as one square.

  FIG. 27 is an image diagram of an example of reception flow data. The response flow data 79 displays whether each procedure includes a silent section or whether the silent section is valid or not by distinguishing it by color coding or the like. For example, in the case of a procedure with a silent section, the reception flow creation unit 67 changes the color of the frame, and fills with a color when the silent section is not valid.

  When there is standard response flow data (standard flow), the response flow data 79 is compared with the standard flow. When there are more procedures than the standard flow, the reception flow creation unit 67, for example, marks many procedures or puts a symbol in the procedure name.

  When the procedure is fewer than the standard flow, the reception flow creation unit 67 marks the procedure of the standard flow that should be in the reception flow data 79 or changes the line, for example. If the order of the standard flow and the procedure is different, the reception flow creation unit 67 puts a symbol in the procedure name of the different procedure, for example.

  The reception flow creation unit 67 calculates the processing time of each procedure from the procedure extraction data 78, calculates the total processing time by adding the processing times of each procedure, and generates the response procedure time information 80. FIG. 28 is an explanatory diagram showing an example of reception procedure time information. The response procedure time information 80 in FIG. 28 is a set of procedure ID, time, presence / absence of silent section, and validity of silent section.

  The response procedure time information 80 indicates the time for each procedure, the presence / absence of a silent section, and the validity of a silent section. The response procedure time information 80 represents the total processing time and the presence / absence of a silent section (OR of the presence / absence of a silent section of all procedures).

  The reception flow creation unit 67 outputs the created reception flow data 79 and reception procedure time information 80. That is, the reception flow creation unit 67 outputs reception flow data 79 representing the explanation procedure extracted from the call recording data 71 and reception procedure time information 80 representing the time required for each procedure and the total time required for the explanation phase. To do.

(Creation of voice recognition keyword list 75 and procedure keyword table 77)
FIG. 29 is an explanatory diagram showing a procedure for creating a voice recognition keyword list and a procedure keyword table. The procedure keyword extraction unit 100 extracts important phrases from manuals, FAQs, procedure manuals, and the like, and associates them with each procedure.

  FIG. 30 is an image diagram showing processing performed by the procedure keyword extraction unit. Note that a keyword extraction method from text, which is a well-known technique, is used to determine an important phrase. In the case of FIG. 30, “program disk”, “double click”, “folder”, and the like are extracted as important words. The procedure keyword extraction unit 100 associates the extracted important phrases with procedures described in manuals, FAQs, procedure manuals, and the like. The procedure keyword extraction unit 100 associates an agent operation performed in the procedure. The procedure keyword extraction unit 100 outputs a procedure keyword table 77.

  FIG. 31 is an image diagram showing processing performed by the corresponding phrase creating unit. The corresponding phrase creating unit 101 creates a corresponding phrase for the extracted keyword extracted by the procedure keyword extracting unit 100. For example, when the extracted keyword is a noun, the corresponding phrase creating unit 101 sets the same phrase or a detailed phrase of the phrase as a corresponding phrase. When the extracted keyword is a verb, the corresponding phrase creating unit 101 sets the corresponding phrase as a corresponding phrase such as “Yes” or “Done”. The corresponding phrase creating unit 101 writes the created corresponding phrase in the procedure keyword table 77.

  FIG. 32 is an image diagram showing processing performed by the voice recognition keyword list creation unit. The voice recognition keyword list creation unit 102 creates a voice recognition keyword list 75 from the procedure keyword table 77. The speech recognition keyword list creation unit 102 creates the agent speech recognition keyword list 75 from the extracted keywords in the procedure keyword table 77. The speech recognition keyword list creation unit 102 creates a customer speech recognition keyword list 75 from the corresponding words in the procedure keyword table 77. The speech recognition keyword list creation unit 102 registers all the usage forms for the verb. As a response word such as “Yes”, for example, a variation such as “Yes” or “Yes” is registered.

  According to the present embodiment, it becomes easy to extract a problem portion in the reception flow. In other words, a procedure including an invalid silence interval, in other words, a procedure where the agent waits for the customer, and calculating the time of the silence interval so that the agent waits for the customer (procedure) And that time. If there is a standard flow, a procedure different from the standard flow can be found immediately.

  In addition, according to the present embodiment, it becomes easy to compare response flows for similar inquiries. That is, by using the response flow data 79, it is possible to compare response flows for similar inquiries. Further, by using the response procedure time information 80, the time required for each procedure can be compared. At this time, by referring to the time of the silent section, the influence degree of the silent section can be examined.

  The reception flow creation apparatus 1 of this embodiment can be used as an optimum explanation procedure extraction apparatus. For example, the optimum explanation procedure extracting device extracts a procedure for each inquiry from a customer, extracts a procedure having the shortest time among the procedures, and outputs it as an optimum explanation procedure.

  The optimum explanation procedure extracting device extracts a procedure by the method of the reception flow creating device 1. When there is a silent section by the customer, the optimum explanation procedure extracting device does not include the silent section in the handling time. The optimum explanation procedure extracting device extracts the explanation procedure which is the shortest case for each inquiry, and outputs the explanation procedure which is the shortest case.

  FIG. 33 is a flowchart showing an example of the processing procedure of the optimum explanation procedure extracting device. Proceeding to step S100, the optimum explanation procedure extracting apparatus selects call recording data 71 corresponding to the same inquiry from the response record.

  In step S101, the optimum explanation procedure extracting device generates response flow data 79 for each selected call recording data 71 by the method of the response flow creation device 1. In step S102, the optimum explanation procedure extracting apparatus selects a case having the shortest processing time. Proceeding to step S103, the optimum explanation procedure extracting apparatus outputs a case with the shortest processing time.

The present invention may have the following configurations as described below.
(Appendix 1)
In order to create a response flow from the call recording data recorded from the customer response,
Procedure keywords that are recorded in advance as a procedure keyword table by associating a keyword uttered by the responder to explain the procedure and a keyword uttered by the customer in response to the explanation of the procedure for each procedure of dealing with the customer Table recording means;
Based on prosodic information extracted from the call recording data, utterance interval extraction means for extracting the utterance interval of the customer and the respondent;
Voice recognition means for voice recognition of the call recording data and outputting the recognized keyword, speaker, and appearance time as a voice recognition result;
A procedure extracting means for reading out the procedure keyword table and extracting the procedure corresponding to the recognized keyword included in the speech recognition result based on the procedure keyword table;
Based on the procedure corresponding to the speech segment extracted by the speech segment extraction unit, the speech recognition result output by the speech recognition unit, and the recognized keyword extracted by the procedure extraction unit, the procedure ends from the start. And when there is a silent section other than the speech section, the speaker immediately after the silent section is set as the generator of the silent section, and the time of the silent section is determined from the start to the end of the procedure. A procedure time calculating means for adding to the time;
A reception flow creation program for functioning as reception flow creation means for creating and outputting the reception flow based on the time from the start to the end of the procedure.
(Appendix 2)
The procedure keyword table recording means records the pre-operation or post-operation associated with the procedure in association with the procedure keyword table as the operation contents of the responder for each procedure of the customer response.
When there is a silent section other than the utterance section, the procedure time calculating means performs a pre-work or post-work associated with the procedure before or after the procedure based on the operation content of the responder included in the procedure keyword table. The reception flow creation program according to supplementary note 1, wherein it is determined whether or not there is, and the procedure for adding the silent section is selected based on the determination result.
(Appendix 3)
Based on the operation details of the responder included in the procedure keyword table, if there is a pre-operation or a post-operation associated with the procedure at a position corresponding to the silence interval, the silence interval is determined to be valid. The response flow creation program according to supplementary note 2, wherein if there is no pre-work or post-work associated with the procedure, the silent section is determined to be invalid.
(Appendix 4)
The procedure time calculation means reads the procedure keyword table, the procedure corresponding to the recognized keyword extracted by the procedure extraction means is missing, and the utterance section is at a position corresponding to the missing procedure In this case, the response flow creation program according to any one of appendices 1 to 3, wherein the missing procedure is added according to the utterance period.
(Appendix 5)
The procedure time calculation means includes a recognized keyword included in the voice recognition result output from the voice recognition means, a keyword spoken by the responder to explain the procedure, and a customer uttered in response to the explanation of the procedure. 5. The response flow creation program according to any one of supplementary notes 1 to 4, wherein the procedure including the utterance section is determined based on a keyword.
(Appendix 6)
The computer further estimates the initiative speaker of the speech segment extracted by the speech segment extraction means, and determines the speech segment in which the customer is the initiative speaker immediately after the start of the call recording data. The response flow creation according to any one of appendices 1 to 5, wherein the utterance section after the question phase is set as an answer phase, and the speech recognition by the voice recognition means is functioned as a phase estimation means for limiting to the answer phase. program.
(Appendix 7)
A reception flow creation method in which a computer creates a reception flow from call recording data obtained by recording a reception call to a customer.
A procedure keyword table that is recorded in advance as a procedure keyword table in association with a keyword uttered by the responder for explaining the procedure and a keyword uttered by the customer in response to the explanation of the procedure for each procedure of dealing with the customer Recording step;
Based on the prosodic information extracted from the call recording data, the utterance interval extraction step of extracting the utterance interval of the customer and the responder;
A voice recognition step of voice-recognizing the call recording data and outputting the recognized keyword, speaker, and appearance time as a voice recognition result;
A procedure extracting step of reading out the procedure keyword table and extracting the procedure corresponding to the recognized keyword included in the speech recognition result based on the procedure keyword table;
Based on the procedure corresponding to the speech segment extracted in the speech segment extraction step, the speech recognition result output in the speech recognition step, and the recognized keyword extracted in the procedure extraction step And when there is a silent section other than the speech section, the speaker immediately after the silent section is set as the generator of the silent section, and the time of the silent section is determined from the start to the end of the procedure. A procedure time calculation step to add to the time;
And a reception flow creation step of creating and outputting the reception flow based on the time from the start to the end of the procedure.
(Appendix 8)
A service flow creation device for creating a service flow from call recording data obtained by recording a service call to a customer,
Procedure keywords that are recorded in advance as a procedure keyword table by associating a keyword uttered by the responder to explain the procedure and a keyword uttered by the customer in response to the explanation of the procedure for each procedure of dealing with the customer Table recording means;
Based on prosodic information extracted from the call recording data, utterance interval extraction means for extracting the utterance interval of the customer and the respondent;
Voice recognition means for voice recognition of the call recording data and outputting the recognized keyword, speaker, and appearance time as a voice recognition result;
A procedure extracting means for reading out the procedure keyword table and extracting the procedure corresponding to the recognized keyword included in the speech recognition result based on the procedure keyword table;
Based on the procedure corresponding to the speech segment extracted by the speech segment extraction unit, the speech recognition result output by the speech recognition unit, and the recognized keyword extracted by the procedure extraction unit, the procedure ends from the start. And when there is a silent section other than the speech section, the speaker immediately after the silent section is set as the generator of the silent section, and the time of the silent section is determined from the start to the end of the procedure. A procedure time calculating means for adding to the time;
A reception flow creation device comprising reception flow creation means for creating and outputting the reception flow based on the time from the start to the end of the procedure.

  The present invention is not limited to the specifically disclosed embodiments, and various modifications and changes can be made without departing from the scope of the claims. For example, if there is an utterance section in which speech recognition has caused an error, it may be estimated by a known technique or ignored.

It is explanatory drawing of an example showing the outline | summary of a present Example. It is an image figure of an example showing the process which specifies the procedure location from the input telephone call recording data, and calculates the required time of each procedure. It is an image figure of an example showing the process when there is a gap among the processes which calculate the required time of each procedure. It is an image figure of an example showing the process with respect to a silence area among the processes which calculate the required time of each procedure. It is an image figure of the other example showing the process with respect to a silence area among the processes which calculate the required time of each procedure. It is a hardware block diagram of an example of a reception flow production apparatus. It is a block diagram showing the relationship between the process part and data of a reception flow production apparatus. It is a flowchart of an example showing the process sequence of the reception flow preparation apparatus. It is explanatory drawing which shows an example of the prosody data of an agent and a customer. It is explanatory drawing which shows an example of an utterance area information of an agent and a customer. It is explanatory drawing which shows an example of phase information. It is explanatory drawing which shows an example of a speech recognition keyword list. It is explanatory drawing which shows an example of a speech recognition result. It is explanatory drawing which shows an example of a procedure keyword table | surface. It is explanatory drawing which shows an example of the procedure extraction data entered by the procedure extraction part. It is a flowchart of an example showing the process sequence of the procedure extraction part. It is a flowchart of an example showing the process sequence which a procedure time calculation part performs. It is explanatory drawing which shows an example of the procedure extraction data in which the start time and the end time were entered by the procedure time calculation part. It is a flowchart of an example showing the process sequence of step S21. It is explanatory drawing which shows an example of the procedure extraction data in which the correction of the duplicate procedure and the insertion omission procedure were inserted by the procedure time calculation part. It is a flowchart of an example showing the process sequence of step S22. It is explanatory drawing which shows an example of a silence area information. It is explanatory drawing which shows an example of the procedure extraction data which performed the process of the silence section. It is an example flowchart showing the process sequence of step S23. It is an example flowchart showing the process sequence of step S53. It is a flowchart of an example showing the process sequence of step S57. It is an image figure of an example of reception flow data. It is explanatory drawing which shows an example of reception procedure time information. It is explanatory drawing which shows the procedure which produces a speech recognition keyword list and a procedure keyword table. It is an image figure showing the process which a procedure keyword extraction part performs. It is an image figure showing the process which a corresponding phrase creation part performs. It is an image figure showing the process which the speech recognition keyword list preparation part performs. It is a flowchart showing an example of the process sequence of the optimal description procedure extraction apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Response flow creation apparatus 51 Input apparatus 52 Output apparatus 53 Drive apparatus 54 Auxiliary storage apparatus 55 Main storage apparatus 56 Arithmetic processing apparatus 57 Interface apparatus 61 Prosodic information extraction part 62 Spoken section extraction part 63 Phase estimation part 64 Speech recognition part 65 Procedure extraction Unit 66 Procedure time calculation unit 67 Response flow creation unit 71 Call recording data 72 Prosody data 73 Speech section information 74 Phase information 75 Speech recognition keyword list 76 Speech recognition result 77 Procedure keyword table 78 Procedure extraction data 79 Response flow data 80 Response procedure time information

Claims (6)

In order to create a response flow from the call recording data recorded from the customer response,
Procedure keywords that are recorded in advance as a procedure keyword table by associating a keyword uttered by the responder to explain the procedure and a keyword uttered by the customer in response to the explanation of the procedure for each procedure of dealing with the customer Table recording means;
Based on prosodic information extracted from the call recording data, utterance interval extraction means for extracting the utterance interval of the customer and the respondent;
Voice recognition means for voice recognition of the call recording data and outputting the recognized keyword, speaker, and appearance time as a voice recognition result;
A procedure extracting means for reading out the procedure keyword table and extracting the procedure corresponding to the recognized keyword included in the speech recognition result based on the procedure keyword table;
Based on the procedure corresponding to the speech segment extracted by the speech segment extraction unit, the speech recognition result output by the speech recognition unit, and the recognized keyword extracted by the procedure extraction unit, the procedure ends from the start. And when there is a silent section other than the speech section, the speaker immediately after the silent section is set as the generator of the silent section, and the time of the silent section is determined from the start to the end of the procedure. A procedure time calculating means for adding to the time;
A reception flow creation program for functioning as reception flow creation means for creating and outputting the reception flow based on the time from the start to the end of the procedure. The procedure keyword table recording means records the pre-operation or post-operation associated with the procedure in association with the procedure keyword table as the operation contents of the responder for each procedure of the customer response.
When there is a silent section other than the utterance section, the procedure time calculating means performs a pre-work or post-work associated with the procedure before or after the procedure based on the operation content of the responder included in the procedure keyword table. The reception flow creation program according to claim 1, wherein it is determined whether or not there is, and the procedure for adding the silent section is selected based on the determination result.   Based on the operation details of the responder included in the procedure keyword table, if there is a pre-operation or a post-operation associated with the procedure at a position corresponding to the silence interval, the silence interval is determined to be valid. 3. The response flow creation program according to claim 2, wherein if there is no pre-work or post-work associated with the procedure, the silent section is determined to be invalid.   The procedure time calculation means reads the procedure keyword table, the procedure corresponding to the recognized keyword extracted by the procedure extraction means is missing, and the utterance section is at a position corresponding to the missing procedure 4. The response flow creation program according to claim 1, wherein the missing procedure is added according to the utterance period. A reception flow creation method in which a computer creates a reception flow from call recording data obtained by recording a reception call to a customer.
A procedure keyword table that is recorded in advance as a procedure keyword table in association with a keyword uttered by the responder for explaining the procedure and a keyword uttered by the customer in response to the explanation of the procedure for each procedure of dealing with the customer Recording step;
Based on the prosodic information extracted from the call recording data, the utterance interval extraction step of extracting the utterance interval of the customer and the responder;
A voice recognition step of voice-recognizing the call recording data and outputting the recognized keyword, speaker, and appearance time as a voice recognition result;
A procedure extracting step of reading out the procedure keyword table and extracting the procedure corresponding to the recognized keyword included in the speech recognition result based on the procedure keyword table;
Based on the procedure corresponding to the speech segment extracted in the speech segment extraction step, the speech recognition result output in the speech recognition step, and the recognized keyword extracted in the procedure extraction step And when there is a silent section other than the speech section, the speaker immediately after the silent section is set as the generator of the silent section, and the time of the silent section is determined from the start to the end of the procedure. A procedure time calculation step to add to the time;
And a reception flow creation step of creating and outputting the reception flow based on the time from the start to the end of the procedure. A service flow creation device for creating a service flow from call recording data obtained by recording a service call to a customer,
Procedure keywords that are recorded in advance as a procedure keyword table by associating a keyword uttered by the responder to explain the procedure and a keyword uttered by the customer in response to the explanation of the procedure for each procedure of dealing with the customer Table recording means;
Based on prosodic information extracted from the call recording data, utterance interval extraction means for extracting the utterance interval of the customer and the respondent;
Voice recognition means for voice recognition of the call recording data and outputting the recognized keyword, speaker, and appearance time as a voice recognition result;
A procedure extracting means for reading out the procedure keyword table and extracting the procedure corresponding to the recognized keyword included in the speech recognition result based on the procedure keyword table;
Based on the procedure corresponding to the speech segment extracted by the speech segment extraction unit, the speech recognition result output by the speech recognition unit, and the recognized keyword extracted by the procedure extraction unit, the procedure ends from the start. And when there is a silent section other than the speech section, the speaker immediately after the silent section is set as the generator of the silent section, and the time of the silent section is determined from the start to the end of the procedure. A procedure time calculating means for adding to the time;
A reception flow creation device comprising reception flow creation means for creating and outputting the reception flow based on the time from the start to the end of the procedure.

Images