JP5092720B2 - Product description content transcription support program, apparatus, and method. - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To allow a batch definition of question types of respective sentences by sentence labels defined with the plurality of question types, and to improve the accuracy of data. <P>SOLUTION: When a further specific field is aimed at and performing documentation based on a conversation record between an exponent and a customer, customer situation is grasped; explanation content (assumed explanation definition body) according to a contract purpose is decided, a commodity knowledge level, and experience years, or the like, are determined and explained; and the degree of similarity between actual explanation and assumed explanation is obtained for each process. When the similarity degree is a prescribed degree or lower, the transcription range of conversation record data is reduced, and the number of transcribing persons is increased, according to the similarity degree, and processing for performing a request is performed to aim the accuracy improvement of the documentation. <P>COPYRIGHT: (C)2009,JPO&amp;INPIT

Description

  When conducting verbal explanations of important matters and consent confirmations in operations from finance, insurance, general sales and other sales approaches, product explanations to contract conclusions, there are no problems with the explanations and consent confirmation procedures. This is related to transcription of audio text to confirm whether or not.

Explanation that added new judgment standard (principle of conformity) that does not make solicitation deemed inappropriate in light of the customer's knowledge, experience, property situation and purpose of concluding a contract with the revision of the Financial Instruments and Exchange Act The need for accountability to customers has become increasingly important.
As a method for confirming the contents of the explanation, in recent years, in addition to the recording / recording being explained, the conversation contents are converted into a character string, stored in a time series, and the data is searched and confirmed when necessary.
One patent document related to the above was detected.
The corresponding document has a plurality of speech recognition means and performs independent speech recognition while referring to basic data (dictionary data including speech, language, etc.) in the field where conversation contents can be assumed to some extent. , Voice recognition with high accuracy can be performed (eg, since the voice recognition between the customer and the call center support person is performed independently, even if the voice is superimposed, the voice of the person is increased without being harmed by others. It is emphasized that it can be recognized with accuracy. Then, the characters obtained from the speech recognition result are displayed in parallel (for example, in two steps) on a display device or the like, and the character string displayed in parallel (for example, the conversation content between the customer and the call center supporter is displayed as characters. Since the conversation content is converted into text (compared to the act in which the caller listens to the speech and transcribes the text) while the transcriber is referring to the text (displayed in two rows as a column), the text content can be made more efficient. It is stated.
JP 2005-126869 A

  However, in the technique disclosed in Non-Patent Document 1, only type identification is performed to extract which sentence is a substantial question sentence (nuclear sentence) from the types used in question type identification. Only one type is identified, and a plurality of types are not identified as the subject of the present invention.

  As described above, the question type identification of a natural sentence does not mean that the question sentence is accurately identified by one type identification.

  For this reason, like this non-patent document 1, only one type identification can be performed at a time, and in order to perform a plurality of types of type identification, it is necessary to perform type identification individually for each type. There will be time and resources.

In order to achieve the above-described conventional problems, the present invention provides a plurality of texts provided in sentence label units indicating a plurality of question types including a question type indicating whether or not a sentence indicates a question and storing definitions of the sentence labels. -Bull, storage means, reception function for receiving question information by natural sentence, and the question information received by the reception function are divided into sentence units, and morphological analysis is performed for each sentence to determine the attribute of each word. The morpheme analysis function for determining and storing the result in the storage means, and for each sentence stored in the storage means by the morpheme analysis function, a score by a plurality of tables is obtained, and this score is used. The sentence type of each sentence is determined and stored in the storage means. Based on the sentence label determined by the question type determination function, the question unit is determined from each sentence stored in the storage means. Chan to perform chunking Those having a bridging functions.

  With this configuration, the question type of each sentence can be collectively defined by a sentence label in which a plurality of question types are defined, so that a plurality of question types can be determined at a higher speed than before.

  In addition, by performing chunking, it is possible to cause the questioner to output a more accurate answer than the answer to each sentence unit question.

  FIG. 1 is an example of a system configuration diagram according to an embodiment of the present invention.

  As clearly shown in FIG. 1, the system according to the present embodiment includes a product description content characterizing support device 1 and n client devices 21 to 2n arranged in a sales office 2, m for a transcriber. The client devices 31 to 3m, the k client devices 41 to 4k for editors, and the s client devices 51 to 5s for compliance officers are connected by a dedicated network.

  Among them, the product description content characterizing support device 1 stores the conversation data of the explainer and the customer at the time of product description collected from the client device arranged at the sales office, and distributes the transcription data to the client device for the transcriber. The data collected by the transcriber is collected. In addition, it sends the transcription data to the editor, collects the transcription data modified by the editor, and ultimately controls to pass the clean transcription data to the compliance officer. .

  Further, the client devices 21 to 2n arranged in the sales office are devices that acquire and store conversation data between the presenter and the customer at the time of the product description.

  Further, the transcriber client devices 31 to 3m are devices that reproduce the conversation data of the presenter and the customer and convert the conversation contents into text data.

  In addition, the client devices for editors 41 to 4k are devices that inspect the conversation contents that have been converted into text.

  Next, functional blocks in each apparatus will be described with reference to FIG. FIG. 2 is a functional block diagram of the system in z and z1.

  The client apparatuses 21 to 2n arranged at the branch offices each have a microphone 210 for acquiring sound for the presenter, a voice recording unit 211 for acquiring voice information input to the microphone 210, and a voice recorded by the voice recording unit 211. Voice recognition means 212 for recognizing voice, voice recording means 213 for obtaining the speech recognition result of the explainer output from the voice recognition means 212, microphone 220 for customer voice acquisition, and continuous voice input from the microphone 220 Voice recording means 221, voice recognition means 222 for recognizing the voice recorded by the voice recording means 221, voice recording means 223 for recording the voice recognition result of the customer recognized by the voice recognition means 222, and the explainer's keyboard From a user interface (not shown) such as a voice recording unit 213 Input and display means 224 for displaying the speech recognition result keyword or the like recorded in the sound recording unit 223 and, control each of the means and networks - consisting processing control and communication unit 225 for performing communication control with the click.

The product description content characterization support device 1 responds to the contract purpose and product knowledge level based on the information obtained by inquiring the customer in order to grasp the customer status such as the customer's property status, contract purpose, product knowledge, etc. Customer explanation content presenting means 101 for determining the explanation contents and presenting the explanation contents to the corresponding sales office arrangement client device, and recording information such as product explanation and voice at the time of contract from the client arrangements 21 to 2n at the sales office network. Explanation content collecting means 102 that collects data via the client, transcription data distribution means 103 that distributes recording data such as speech obtained by the explanation content collection means 102 to each of the transcription client devices 31 to 3m, and each transcription Transcript data collected by each client 31 to 3m from the transcription data created by each client device 31 to 3m. The data collection means 104, the transcription data collected by the data collection means 104, the transcription edit data creation means 105 that generates the data for passing to the editor, and the transcription data generated by the transcription edit data generation means 105 Edit data presenting means 106 to be transmitted to the editor client devices 41 to 4k via the client, and the transcription data corrected by the editor is obtained from the editor client devices 41 to 41k via the network. Editing data collecting means 107, and processing control / communication means 108 for controlling communication with each means of the product description contents characterizing support apparatus 1 and the network.

  The client device 31 to 3m of the transcriber includes a speaker 300 for reproducing the voice of the presenter or the customer, an audio reproducing means 301 for reproducing an audio recording and outputting it from the speaker 300, an operator's keyboard, etc. The input / display means 302 for displaying the input from a user interface (not shown), the speech recognition result keyword, etc., the communication means 303, and the processing control means 303 for controlling each means. In addition, although not shown in the figure, a microphone or voice recognition means for acquiring the voice of the operator may be added as a complementary function of the input / display means 302, and an alternative input for the keyboard may be provided by voice input. .

  The client devices 41 to 4k of the editor include a speaker 400 for reproducing the voice of the presenter or the customer, a voice reproducing unit 401 for reproducing the voice recording and outputting it from the speaker 400, an operator keyboard, and the like. Input / display means 402 for displaying the input from the input and speech recognition result keywords, communication means 403, and processing control means 401 for controlling each means. Further, although not shown in the figure, a microphone or voice recognition means for acquiring the voice of the operator may be added as a complementary function of the input / display means 402, and an alternative input for the keyboard may be provided by voice input. .

  The client devices 51 to 5s of the person in charge of compliance include a speaker 500 for reproducing the voice of the presenter or the customer, a voice reproducing means 501 for reproducing the voice recording and outputting it to the speaker 500, an operator's keyboard, etc. Input / display means 502 for displaying the input from the input and speech recognition result keyword, communication means 503, and processing control means 504 for controlling each means. In addition, although not shown in the figure, a microphone or a voice recognition unit for acquiring the operator's voice may be added as a complementary function of the input display unit 502, and an alternative input of the keyboard may be provided by voice input.

  In the system configured as described above, the processing will be described below.

  Note that each process is performed by a corresponding unit of each apparatus.

  First, processing between the product description content characterizing support processing apparatus 1 and each client will be described with reference to the flowcharts of FIGS.

  3 and 4, the dotted frame represents that the operator has operated the client, the product description content characterizing support device, or each client.

  In addition, in this explanation, the work up to the completion of the transcription is within the scope of the essential explanation, and therefore explanation concerning the compliance officer is omitted.

  First, basic processing of the product description content characterizing support processing device 1 will be described.

First, the process control / communication means 108 of the product description content characterization support processing apparatus 1 is activated to prepare for processing. (301)
The process control / communication means 108 waits for a process request coming from each client (302). Here, the presence / absence of a processing request from the client is checked (303), and if not, the process returns to 302, and if present, the process proceeds to 304.

  Next, the process control / communication means 108 of the product description content characterization support processing device 1 analyzes the process request request from each client, and the product description content is also transferred to the corresponding processing point in the home support processing device 1, Each process is executed (304). In the following, there is a description that processing has been performed in response to various requests from each client device. However, as a result of performing the above processing and analyzing the request, the processing is transferred to the corresponding means ( Processing points).

  When the processing for the request from each client ends, the process proceeds to 302 (305).

  Next, basic processing of each client device will be described.

  First, the processing between the branch office client apparatuses 21 to 2n and the product description content characterization support processing apparatus 1 will be described based on the branch office client apparatus 21. FIG.

First, after the sales office client device 21 is turned on, the processing control / communication means 225 in the sales office client 21 is activated to prepare for processing. (308)
Here, if necessary, the presenter depresses the item explanation start key (309 / in the explanation, the process start is defined as the key depression).

  The branch office client device 21 detects that the product description start key is pressed (310). If the key is pressed (311), if the key is detected, the process proceeds to 312 and no key is pressed. The process proceeds to case 309.

  Next, the branch office client device 21 displays an input screen (FIG. 4) for inquiring personal information, property status, product contract purpose, desired purchase product, product knowledge, years of investment experience, etc., and prompts input ( 312).

  Here, it is assumed that the presenter makes an inquiry to the customer, inputs the customer status, and then presses the transmission key (313).

  In response to the detection of the transmission key being pressed, the branch office client device 21 transmits information on the contents of the screen on which the customer status has been input to the product description content characterizing support processing device 1 (313). The information at this time is referred to as customer explanation content presentation request information.

  The product description content characterizing support processing device 1 that has received this information analyzes the processing request request by receiving information from each client, and if it is the customer description content presentation request information as a result of the analysis, the process proceeds to 315. To do.

Here, the product description content presenting means 101 is stored in its own database (not shown), from the customer's information, the customer's property status, product contract purpose, desired purchase product, product knowledge, years of investment experience, Comparing other input information with the customer explanation statistical information file accumulated from the past, etc., and a predetermined extraction method (for example, extracting by the degree of relevance of the two factors of allowable risk degree and investment experience degree desired by the customer) The description content assumed to be the best (FIG. 5 assumed explanation definition body) is determined and returned to the branch office client apparatus 21 that has requested the assumed explanation definition body (315).

  Here, the store client device 21 receives the assumed explanation definition body, and the processing control / communication means 225 includes a timer (installed in the sales client device 21, not shown), a microphone 210, a voice recording unit 211, a voice recognition unit. The apparatus and processing of the means 212, the voice recording means 213, the microphone 220, the voice recording means 221, the voice recognition means 222, and the voice recording means 223 are activated to start voice recording and prepare for voice recording (FIG. 4, 316).

The microphone 210, the voice recording means 211, the voice recognition means 212, and the voice recording means 213 are for the explainer, and the microphone 220, the voice recording means 221, the voice recognition means 222, and the voice recording means 223 are for the customer.
The data output by the voice recording unit 211 and the voice recording unit 221 records the continuous voices of the explainer / customer in time series and saves them in a file or the like. The voice recording unit 213 and the voice recording unit 223 record voice data uttered by the presenter or the customer, voice data recognized and output by the voice recognition unit 212 and the voice recognition unit 222, and data such as converted text in time series. Add saving to file etc. Details of this process will be described later in the process of the process control / communication means.

  Here, when the instructor performs the product description end operation at the end of the product description (317), the process control / communication unit 225 that detects the operation performs the timer, the microphone 210, the voice recording unit 211, the voice recognition unit 212, and the voice recording. The processing operations of the means 213, the microphone 220, the voice recording means 221, the voice recognition means 222, and the voice recording means 223 are stopped, and the recording processing of data such as voice is terminated (318).

  Then, the sales client device 21 edits the recorded voice data and the like, and transmits it to the product description content characterizing support processing device 1 by the processing control / communication means 225. The processing request at this time is an explanation content collection request (319).

  Further, after the process of 319 is completed, the branch office client apparatus 21 proceeds to the process of FIG. 3/309 after the transmission is completed normally, and executes the termination process and terminates at the termination.

  Upon receiving the explanation content collection request from each of the branch office clients 21 to 2n transmitted as described above, the product explanation content characterization support apparatus 1 generates data based on the received collection data of the explanation content, and the identification name Is created, a folder is created with the corresponding identification name, and a file is saved for each data such as the voice of the explainer or customer. By this saving, the data to be transcribed is in the existing state (321).

  If the presence or absence of the transcription data is checked, a message requesting the transcription is sent to the corresponding transcription person (504 in FIG. 5). Thereafter, the processing shifts to the return destination of the caller (305 in FIG. 3).

  Next, the processing of the transcription client devices 31 to 3m will be described using the transcription client device 31 as an example.

First, after the transcription client device 31 is powered on, the transcription client processing control means 303 is activated to prepare for processing (FIG. 5, 501).

  Here, when the transcriber operates the transcribe client device 31 and receives a message for a transcribe request from the product description content characterizing support device 1 or performs an inquiry operation on the presence of work, the transcribe client device. 31. When the process control means 303 performs the process and displays the result, the transcriber knows whether or not the work is present and, if the work can be performed, transcribes the requested operation for the transcribed data and the client device. 31 (503).

  In response to this operation, the transcription client device 31 transmits a processing request request as a transcription data distribution request by a data request operation. At this time, the identification number of the transcriber is also added and transmitted to the product description content characterizing support apparatus 1.

  Upon receiving this transcription data distribution request, the transcription data distribution means 103 of the product description content characterizing support apparatus 1 extracts the transcription data corresponding to the skill level of the transcription person and corresponds to the transcription client apparatuses 31 to 31. Distribute to 3m (505). Thereafter, the process returns to the caller's return destination (FIG. 3, 305). Details of the transcription data distribution means 103 will be described later.

  When the data distributed at 505 is transcribed and received by the processing control means 303 of the client 31, the received transcribed data is reproduced and the transcriber who hears the data performs data input processing (507). The processing of the processing control unit 304 at this time will be described later (508). Thereafter, when an input for instructing the end of work is detected from the operator, the process control means 304 and the communication means 303 transmit the transcription data to the product description content characterizing support apparatus 1 (509). The processing request is a transcription data collection request.

  Further, the transcription client device 31 normally proceeds to (502) after the transmission is completed. At the time of termination, termination processing is executed and terminated.

  Upon receiving this information, the transcription data collection means 104 of the product description content characterization support apparatus 1 transcribes the received transcription data and stores the data in the data collection file (FIG. 6, 511). This process is performed by the transcription data collection means 104, and details of this process will be described later.

  Next, the transcript editing data creation means 105 of the product description content characterization support apparatus 1 checks both the explainer and the customer data management unit corresponding to the collected transcript data, and all the middle process processing steps. If the status is “2” (all collected), an edit data presentation file for the transcription editor is generated (512). This processing is performed by the transcription editing data creation means 106, and details of this processing will be described later.

  Further, if the presence / absence of the edit data presentation file is checked, a message is sent to the editor and a request for editing is sent to the editor. (FIG. 7: 513) Thereafter, the process returns to the caller's return destination.

  It is assumed that after the editing client device is turned on, the editing client processing control means 41 is activated and ready for processing. (514) In the present embodiment, the process of the editor client 41 is taken as an example, but the other editor clients 42 to 4k can perform the same process.

When the editor operates the editor client 41 and receives an edit request message or knows the presence or absence of the editing work by the inquiry instruction of the work presence, if the work can be performed, the editor client 41 is further opened. Operate the corresponding edit data presentation request. (515)
Upon receiving this edit data presentation request, the editing client 41 transmits the processing request request as an edit data presentation request. (516)
Upon receiving this edit data presentation request information, the product description content characterization support device 1 shifts the processing to the edit data presentation 106 (517), refers to the edit data presentation file, and sends the corresponding edit data to the editor client. It returns to 41 and returns to the return destination of the caller. (518)
Upon receiving this information, the processing control means 404 of the editor client 41 (519) displays the received download data, and executes this data editing processing in accordance with the editor's operation. (520: Details will be described in the processing of the processing control unit of the editing client described later.)
When an operation indicating the end of work by the editor is received, the data is edited by the communication means 403 and transmitted to the product description content characterizing support apparatus 1. (521: The processing request at this time is the editing data collection request. (521).

  Note that the editor client device 41 normally shifts the process to (515) after the transmission is completed, and terminates the termination process when terminated.

  Next, upon receiving this edit data collection request from the editor client 41, the product content explanation apparatus 1 that has received this edit data collection request shifts the processing to the edit data collection means 107 (522).

  The edit data collection unit 107 generates edit data collection files collectively based on the corresponding edit data and saves the files (523).

  Thereafter, the process returns to the caller's return destination. Thus, the file is saved and the transcription data is completed. In addition, the edited data is handed over to the compliance person who is the client.

  FIG. 8 shows an example of the personal information input screen described above, based on account number, personal information, property status, product contract purpose, desired product or acceptable risk level, investment experience, product knowledge level, etc. Become.

  Based on this input information, a customer explanation statistical information file or the like is searched, and the explanation content (assumed explanation definition body) assumed to be the best according to the customer situation is determined.

  Next, the original assumed explanation definition body information will be described with reference to FIG.

  The assumed explanation definition body information is a data table for product explanation created for the purpose of proceeding with explanation without fail without fail in order to fulfill accountability, and storage means (Fig. (Not shown).

  This assumed explanation definition body information is divided into three types, and is created according to the customer's product knowledge level etc. (for example, beginner, intermediate, advanced), and results data accumulated in statistical information Based on the above, a large number are prepared for each product at each customer level.

One assumed explanation definition body is composed of a plurality of large processes, and the large process is composed of a plurality of intermediate processes. One middle process is an evaluation unit for text transcription, and is composed of a plurality of small processes, and a data string of the small processes is a minimum unit. The data length from the start to the end of the process is arbitrary.
Processes can be freely set as large processes, medium processes, and small processes. However, an identification number or the like is assigned to the process so that the process can be identified.

In this embodiment,
Examples of large process identification numbers: 1, 2, 3, ...
Middle process identification number assignment example: 1-1, 1-2, 1-3, ...
Small process identification number assignment example: 1-1-1, 1-1-2, 1-1-3,...
And so on.

Next, the structure of the definition body of the middle process and the small process will be described based on the definition body diagrams of FIGS.
Small process definition body is progress small process, explainer / customer classification, story text, estimated time, actual start time (elapsed time), explainer registered keyword, utterance keyword, unexpected keyword, small process similarity, customer Registered keywords, utterance keywords, unexpected keywords, and small process similarity, the middle process definition body adds the middle process similarity degree of the explainer and the middle process similarity degree of the customer to the set of small process definition bodies Consists of items.
The progress small process is a story consisting of a given small process identification number, the explanation / customer category is explanation: 0, customer: 1 is the identifier, and the story text is a story consisting of the character strings of the explanation and the customer. -When the definition is created, the assumed time is the assumed conversation time of the explainer or customer, and the registered keyword of the explainer or customer is a keyword that expresses the voice of the explainer or customer as a character string. To do.

  The actual start time (elapsed time) is measured and stored in the area of the corresponding item after the start time and elapsed time of the actual conversation of the presenter or customer.

  Explainer or customer utterance keywords (character strings output by voice recognition) are applicable each time a character string that matches a registered keyword (set in the assumed explanation definition body) of the explainer or customer appears. Is stored in the area portion.

  The unexpected keyword of the explainer or customer is stored in the area portion of the corresponding item whenever a character string that does not match the registered keyword of the explainer or customer appears.

The degree of similarity of the small process of the explainer or customer is obtained and stored by the degree of similarity calculation method. (For details, refer to the explanation of the similarity extraction matrix for each product below)
The intermediate process similarity of the explainer or customer is stored by obtaining the lowest numerical value in each of the small process similarities.

  Next, the product similarity extraction matrix will be described with reference to FIG.

  In this commodity similarity extraction matrix, the relationship between the keyword match rate d and the mismatch rate (e) is arranged in a matrix, and the numerical values generated from the similarity are placed in each matrix. Based on this intersection, a value relating to the similarity is extracted.

Each element shown in FIG. 12 is as follows.
a: Number of registered keywords set by the explainer or customer However, when the attribute of the corresponding keyword is a synonym, the number is not counted.
b: Number of keywords spoken by the explainer or customer c: Number of keywords unexpected by the explainer or customer d: Keyword match rate = b ÷ a
e: Keyword mismatch rate = c ÷ (b + c)
In this way, the relationship between the keyword match rate (d) and the keyword mismatch rate (e) is arranged in a matrix, and numerical values born from an empirical rule that refers to the intersection (intersection area) between d and e are stored. The corresponding numerical value is extracted, and the degree of similarity for each process is obtained.

  Incidentally, although the numerical value of the similarity degree is not particularly specified, in this embodiment, the similarity degree numerical value is obtained when the keyword mismatch rate (e) is higher than the keyword matching rate (d) as indicated by the line (a) in FIG. It is assumed that there is a downward trend.

Further, the difference (absolute value) between the assumed explanation time and the actual time is obtained, and when the difference value is greater than or equal to a predetermined value, the degree of similarity is obtained by multiplying by a coefficient that reduces the degree of similarity degree according to the increase of the difference value.
Incidentally, although the numerical value of the degree of similarity is not particularly specified, in the present embodiment, it is assumed that the tendency is like a line (A) in FIG.

Next, the details of the definition of registered keywords and the function description of FIG. 13 will be described.
The definition body of the registered keyword of the explainer or the customer is composed of a registered keyword number, keyword wording, next process progress, presence / absence of synonyms, wording continuation presence / absence, and operation command sequence.
The registered keyword number consists of 5 characters. When the first character is S, it is an explainer, when it is K, a customer is indicated, and thereafter, 4 characters indicate a processing serial number.
In the keyword wording, the first two characters represent the number of characters of the keyword wording, and the subsequent character string represents the keyword wording.
The progress of the next process represents 1: a keyword for proceeding to the next process, and 0: a keyword for the current process.
This attribute exists only for the presenter.
The synonym presence / absence indicates whether the keyword attribute is a synonym: 1: synonym, 0: non-synonym, and the registration keyword parameter is obtained in the process of obtaining the degree of similarity. Not subject to keyword parameter counting.
The word continuation presence / absence indicates 1: continuation exists, 0: no continuation, and is used to determine whether or not the keyword group is the last.
In the operation command string, the first two characters represent the number of characters of the entire operation command string, and the following three characters represent a unique operation command, and the operation command description is repeated as necessary.
For example, the registered keyword number in the first line of the registered keyword definition: “S0001”, the keyword wording: “27 Then I will proceed to risk explanation regarding the next exchange”, the next process progress: “1”, presence of synonyms : “0”, word continuation presence / absence: “1”, operation command string: “06A01B02”, the corresponding variable of the number of utterances is initialized by the operation command first “A01”, and the number of utterances is incremented by one. , The utterance order (number of utterances): “1” is obtained, and by the next operation command “B02”, the start time at the time of utterance, the elapsed time, the utterance order (number of times), the voice data matching the registered keyword, Since the recording of the character string data is instructed, the utterance time: HH: MM: SS: ZZZ (hour minute second millisecond), elapsed time: MM: SS: ZZZ (minute second millisecond) Voice order: 1, Voice data: “Now let me move on to explain the risk of the next exchange”, Character string: “Let me move on to explain the risk of the next exchange”, explained by the voice recording means 213 The voice data of the presenter is recorded in the presenter voice recording file 11 in time series.

  FIG. 14 is a flowchart of the processing control / communication means 225 of the sales office client, which will be described in the order of processing. In the following description, the process number indicates the process table information number shown in FIGS. 10 and 11.

The outline of this processing is when the character string data obtained through the speech recognition means 212 or 222 is compared with the character string registered for each small process of the assumed explanation definition body in the conversation between the presenter and the customer, and they match. In other words, information such as corresponding voice data and character strings is stored (recorded) in the utterance keyword item area. Further, in the case of mismatch, information such as corresponding voice data and character strings is stored (recorded) in an unexpected keyword item area.
The progress of the process is controlled by performing speech recognition of words (uttered keywords) uttered by the presenter and collating this with the process chart.

  For example, after the explanation of the corresponding process (02-05-01) shown in FIG. 10 is completed and the customer's understanding is confirmed, a progress keyword for proceeding to the next process (02-05-02) is spoken. And so on. Specifically, the processing control communication unit 225 of the sales office client executes the following processing.

  In executing this processing, initialization of variables and areas, initial processing, and the like are performed (1301). For example, there is a process of taking out the value of the progress small process set at the head of the assumed explanation definition body and storing it in the progress process variable.

Next, the voice recognition unit 212 and the voice recognition unit 222 are activated, and the voice recognition operation state for the speech of the presenter or the customer is set. (1302)
Next, the input display means 10 is activated, and an input such as a keyboard and a display device are put into operation. (1303)
An example of the screen display on the explainer side at this time is shown in FIG.

Next, the timer, the microphone 10, the voice recording unit 211, the microphone 20, and the voice recording unit 20 are activated and the recording of the data such as the voice of the presenter / customer is started. (1304)
The apparatus and process of the voice recording unit 213 and the voice recording unit 21 are activated, preparation for starting recording of voice and the like is performed, and an operation instruction from the process control / communication unit 10a is awaited. (1305)
Confirms the presence of events such as input from the keyboard, speech recognition output of utterances, and operation commands. (1306/1307), go to (1306) when there is no event. When there is an event, the process proceeds to (1308).

When an event occurs, the event type is checked and distributed to each process. (1308)
In the case of the end notification of the assumed explanation definition body, the process proceeds to FIG. 16: 1325 (FIG. 14: 1309). In the case of an input notification such as a keyboard, the process proceeds to FIG. 16: 1326 (FIG. 14: 1310). In the case of voice recognition output, the processing shifts to FIG. 15: 1309 (FIG. 14: 1311). Otherwise, it is determined as an error, and the process proceeds to (FIG. 14: 1312).

  Incidentally, the end notification of the assumed explanation definition body is generated by executing the operation command of the “notification of expected explanation definition body end” described in the final registered keyword definition body.

  Note that the processing of FIG. 14: 1312 executes error processing and shifts the processing to 1306.

Next, if it is determined in FIG. 14: 1311 that the process proceeds to FIG. 15: 1309, the channel of the voice recognition output information is checked, and if the output source is the explainer side in the voice recognition output channel variable, “0”, customer In the case of the side, “1” is set. (Fig. 15: 1309)
Then, the corresponding speech text (utterance keyword) is extracted from the speech recognition output information (FIG. 15: 1310).

  Further, the voice recognition output channel variable is checked, and if the presenter is “0”, the process proceeds to 1316, and if the customer is “1”, the process proceeds to 1312 (FIG. 15: 1311, FIG. 15: 1312). .

Next, the customer registration keyword group of the definition body of the progress small process identification number stored in the progress process variable is extracted (1313).

  Thereafter, it is checked whether or not there is a character string that matches the utterance keyword from the extracted registered keyword group. If there is a matching character string, the process proceeds to (FIG. 16: 1329), and if there is no matching character string, the process proceeds to FIG. 16: 1326. Transition (1314, 1315).

  When the process is shifted to 1316, the explainer registration keyword group of the definition body of the progress small process identification number stored in the progress process variable is extracted (1316).

  Then, the presence / absence of a character string that matches the utterance keyword is checked from the extracted registered keyword group. If there is a matching character string, the process proceeds to 1329, and if there is no matching character string, the process proceeds to 1319 (1317). , 1318).

When the process shifts to 1319, the explainer registration keyword group of the definition body of the small process identification number next to the progress small process identification number stored in the progress process variable is extracted. (1319)
Next, it is checked whether or not there is a character string that matches the utterance keyword from the extracted registered keyword group. If there is no matching character string, the process proceeds to 1325 (1320, 1322).

  When the process shifts to 1323, it is checked whether or not it is a next process progress target keyword, and in the case of the next process progress target keyword, the process proceeds to FIG. 16: 1328. In the case of the next process progress target keyword, the processing shifts to FIG. 16: 1325 (1323, 1324).

  A message to the effect that the voice has not been matched with the keyword group generated in the current or next progress process is displayed, and if necessary, a retry by a recurrent voice or a progress process change operation by keyboard input is requested (1325). This message attribute is a message that is unilaterally notified to the presenter and does not require the presenter's response to the message.

Next, the voice recording unit 213 or 21 of the corresponding channel is instructed to record the voice data of the utterance keyword, the start time at the time of utterance, the elapsed time, and the utterance order. Thereafter, the data is recorded in time series by the corresponding recording means 11 or 21. (1326)
Furthermore, the start address of the explainer or the customer's progress small process is obtained from the progress small process identification number stored in the progress process variable and the identifier stored in the voice recognition output channel variable, and then the corresponding actual start time, elapsed time Each address in the expected definition body of the time and the unexpected keyword is obtained, and the start time at the time of utterance, the elapsed time, and the utterance order are added to the unexpected keyword and stored in each corresponding area of the definition body (1327), 1306 Transfer processing to.

  Next, when it is determined that the processing is shifted to 1328, since the same keyword as the call generation keyword is detected from the group of explainer registration keywords of the next process identification number, the similarity degree of the current progress process is obtained, and the corresponding similarity degree is obtained. Store in the area. At this time, when changing to a new intermediate process, the lowest numerical similarity in the degree of similarity of the current progress small process is obtained and stored in the corresponding intermediate process similarity area. Thereafter, the progress process variable value is changed to the next process identification number, and the utterance order is initialized.

  In addition, the method of calculating | requiring the similarity degree of a small process is as having specified in the said description.

  Further, the voice recording data of the presenter or customer utterance keyword recorded by the voice recording means 213 or 223 (refers to voice data corresponding to the utterance keyword or unexpected keyword of the presenter or customer) is a small process unit. Are combined into a single audio data and re-stored (1328).

  The action command string defined in the registered keyword definition that matches the character string is executed. This instructs the voice recording means 213 or 21 of the corresponding channel to record the voice data of the utterance keyword, the start time at the time of utterance, the elapsed time, and the utterance order. Thereafter, the data is recorded in time series by the corresponding recording means 11 or 21 (1329).

  Next, the start address of the progress subprocess of the explainer or customer is obtained from the progress subprocess identification number stored in the progress process variable and the identifier stored in the voice recognition output channel variable, and then the corresponding actual start time and elapsed time Obtain each address in the expected definition body of the time and utterance keyword, add the utterance order to the start time, elapsed time, and utterance keyword at the time of utterance, and store in each corresponding area of the definition body (FIG. 17: 1330), The process proceeds to FIG.

  If the process to 1331 has been transferred, a process end confirmation message is displayed to notify the end of the assumed explanation definition body, and the instructor is notified of the fact, and then the process proceeds to FIG. Migrate (1331).

  Further, when the process is shifted to 1332, it is checked whether or not the process is a progress process change operation by input from a keyboard or the like. If the process is a progress process change operation, the process is shifted to 1335. Otherwise, the process proceeds to 1334 (1332/1333).

In 1333, in the case of the progress process change operation, the necessity: Since control by 100% voice is impossible, the manual change operation of the progress process is accepted.
In addition, the contents of the progress process variable are changed according to the explainer operation. At this time, the fact that there has been a manual progress process operation is left as a history, and the process proceeds to 1306 (1334).

If it is determined in 1333 that the operation is not a progress process change operation, it is checked whether the response is a business end response by the explainer. If the response is a business end response by the explainer, the present processing is terminated. (1335, 1336).
If it is determined in 1337 that the other processing is performed, the other processing is performed, and the processing proceeds to 1336.

  Next, FIG. 19 shows a screen display example of the presenter.

As shown in FIG. 19, this screen display is composed of a progress process, an explainer / customer classification, a story text, an explainer's utterance keyword, an unexpected keyword, a customer's utterance keyword, and an unexpected keyword.
The progress process represents the current process position by background color reversal or the like. In the story text, an assumed text of the explainer / customer is displayed. Explainer's utterance keyword and unexpected keyword are converted into text and displayed in the utterance keyword column if they match the registered keyword in the assumed explanation definition body. If they do not match, they are displayed in the unexpected keyword field. Customer utterance keywords and unexpected keywords are also displayed in the same manner.

  FIG. 20 is an example of recorded data of the presenter's actual voice start time, voice elapsed time, utterance keyword, unexpected keyword, and similarity obtained by the process control / communication means 10a, and is expressed in an easy-to-read manner. It is. Actual recording data is stored in the assumed explanation definition body. The customer's record data is not specified but is recorded in the same way.

  Among these, FIG. 20A is an example of the continuous voice data of the explainer or the customer recorded by the voice recording means 211 or 20. One voice is made up of a voice portion and a start time (e.g., t1) and an end time (e.g., t2), which are voiced portions, which are continuously recorded.

  Further, FIG. 20B is an example of recorded data of the utterance keyword unit voice of the explainer or the customer recorded by the voice recording means 213 or 21. One piece of voice data includes a start time (eg, t1) and an elapsed time (eg, t2) and voice data. The relationship between FIG. 20A and FIG. 20B is a relationship in which continuous speech data is divided into utterance keyword units.

  Further, when progressing to a new process, the voice keyword unit voice recording data so far is logically combined into small process units and combined into one voice data.

  FIG. 21 is an example of an explanation content collection file generated by the processing by the explanation content collection means 30 of FIG. 4: 321.

  The first layer is the folder name of the description contents collection data (SYyyyymmddXXXX, SY: collection data identification, yyyymmdd: creation date, XXXX: serial number of the matter), the second layer is the data management unit for the explainer and customer, and assumed explanation definition Body, the folder name for the explainer, and the folder name for the customer.

  The assumed explanation definition body is as shown in FIG. The third hierarchy is composed of the continuous voice recording data of the explainer or customer shown in FIG. 20A, and the name of the voice recording data folder of the explainer or customer keyword unit. The fourth level is the audio recording data file of the explainer or customer keyword unit shown in FIG.

  FIG. 22 is an explanation of the data management unit that stores the explanation of the explainer and the customer at the time of collecting explanation contents in a table.

  This management department manages the progress of work in the middle process for each explainer and customer.In addition to the data size, data identification name, generation date, serial number, last update date, number of middle processes, the number of middle processes ( 1) Structure of identification number, middle process (1) similarity degree, middle process (1) processing status, middle process (1) number of writers, middle process (1) worker ID is middle process (n) It consists of a structure repeated up to the identification number. The description of each item is as shown in FIG.

  Further, FIG. 23 is a flowchart showing the processing of the transcription data distribution means 103, which will be described according to the processing order.

  First, it is checked whether or not the transcription data distribution file based on the corresponding explanation content collection file has been generated. If the file has been generated, the process proceeds to FIG. 26: 2235, and otherwise the process proceeds to 2203 (2201). 2202).

  If it is determined as other in 2202, a folder or file for distribution data is collectively created based on the corresponding explanation content collection file (2203).

  At this time, the folder and file structure of the generated data are as shown in the table shown in FIG. 27, and the data is generated by taking over the contents of the explanation content collection data.

  For the image of the assumed explanation definition body in the generated folders and files, see the example of assigning the small process identification number of the assumed explanation definition body at the time of process reorganization in FIG. However, only the data space is secured for the reorganization progress small process and the explanation input data of the explainer or the customer, which are not in the definition body at the time of explanation content collection.

In addition, in the small process unit audio file name to which the item has been added, the keyword unit audio record data of the explainer or the customer is grouped into one small process unit, and the audio file name is stored (2203).

  Next, the pointer of the assumed explanation definition body for transcription is set to the first progress small process identification number position (2204).

  Next, the small process similarity degree (for the explainer or customer) pointed to by the pointer of the assumed explanation definition body is checked. At this time, if the numeric value is greater than or equal to a predetermined value, the process proceeds to 2207, and otherwise, the process proceeds to 2208 (2205, 2206).

  When an attempt is made to process 2207 in 2206, the contents of the story text of the explainer or customer pointed to by the pointer of the assumed explanation definition body are copied to the explanation input data of the explainer or customer. By making the copied data as input data candidates at the time of transcription, the input burden of transcription is reduced (2207).

  Next, it is checked whether the pointer pointed to by the assumed explanation definition body is final. In this check, when the pointer is final, the process proceeds to 2211, and in other cases, the process proceeds to 2210 (2208, 2209).

  If it is determined in 2209 that it is other, the pointer of the assumed explanation definition body is updated to the next progress small process identification number position, and the process proceeds to 2205 (2210).

  If it is determined at 2209 that the pointer is the final pointer, the pointer of the assumed explanation definition body for transcription is set at the first in-progress process identification number position (2211, FIG. 24: 2212).

  Then, the intermediate process similarity indicated by the pointer of the assumed explanation definition body is extracted (2213).

  Further, the transcription work distribution table is searched with the extracted intermediate process similarity degree, the similarity degree range is determined, the transcription division number is obtained, and when the division number is “0”, the process goes to 2217. In other cases, the processing shifts to 2216 (2214, 2215).

If it is determined in 2215 that the division number is not “0”, the middle process for the explainer or the customer in the data management unit (for transcription data distribution) is reorganized (divided) with the obtained division number. . As a result, the data size, the number of intermediate processes, and related areas of the data management unit (FIG. 29) are also reorganized. In FIG. 30, (1) before the division and (2) after the three division, the state of the reorganization (division) of the middle process in the data management unit is illustrated.
Further, a new reorganization progress small process identification number is assigned to the reorganization progress small process item in the assumed explanation definition body (see FIG. 30).

  Also. When the number of divisions is set to “0” in 2215, it is checked whether the pointer pointed to by the assumed description definition pointer is final. If the pointer is final, the process proceeds to 2220, and the process proceeds to 2219 in other cases. (2217, 2218).

  If it is determined in 2218 that the process shifts to 2219, the pointer of the assumed explanation definition body is updated to the next in-progress process identification number position, and the process shifts to 2213 (2219).

  If it is determined in 2218 that the processing shifts to 2220, the pointer of the assumed explanation definition body for transcription is set to the in-progress ongoing process identification number position (2220).

  Next, the intermediate process similarity indicated by the pointer of the assumed explanation definition body is extracted (2221).

  Here, using the extracted intermediate process similarity value, the transcription work distribution table of FIG. 16 is searched, the similarity degree range is determined, the transcription aggregation number is obtained, and the aggregation number is “0”. The processing shifts to FIG. 25: 2231, and the other processing shifts to FIG. 25: 2224 (2222, 2223).

  When the process shifts to 2224 at 2223, it is checked whether the pointer pointed to by the assumed explanation definition body is final. If the pointer is final, the process shifts to FIG. 26: 2234. The processing is shifted (2224, 2225).

  If the process shifts to 2226 in 2225, the pointer of the assumed explanation definition body is updated to the next progress process identification number position (2227).

  Next, the intermediate process similarity indicated by the pointer of the assumed explanation definition body is extracted (2228).

  Next, the transcription work distribution table of FIG. 16 is searched with the extracted intermediate process similarity degree value, the similarity degree range is determined, the transcription aggregation number is obtained, and the aggregation number is “0”. If not, the process proceeds to 2230 (2228, 2229).

  When the processing shifts to 2230 at 2229, the high degree of similarity is continuously generated, so that the middle process in the data management unit (for transcription data distribution) is reorganized (aggregated). Along with this, the data size, the number of intermediate processes, and related areas of the data management unit (FIG. 29) are also reorganized (2230).

  In FIG. 30, (2) before consolidation and (3) after consolidation, the state of the reorganization (aggregation) of the middle process in the data management unit is shown in the figure.

  Further, the reorganization progress small process item in the assumed explanation definition body is also reorganized to a new reorganization progress small process identification number (see FIG. 31).

  Next, it is checked whether the pointer pointed to by the assumed description definition body is final. If the pointer is final, the processing shifts to FIG. 24: 2234, and otherwise the processing shifts to 2233 (2231, 2232).

  If it is determined in 2232 that the process shifts to 2233, the pointer of the assumed explanation definition body is updated to the next in-progress process identification number position, and the process shifts to FIG. 24: 2221 (2233).

  Also, in the case of processing to 2234 in FIG. 24: 2234, reference is made to the intermediate process similarity in the data management unit after division or aggregation, the transcription work distribution table is searched, and the similarity range is determined. Then, the number of persons engaged in transcription is obtained, stored in the area for the number of persons engaged, the space corresponding to the number of persons engaged is secured, the reorganization for all the middle processes is executed, and the reorganization of the data management unit is completed ( 2234). FIG. 31 shows an example of the data management unit after the reorganization.

Next, the transcriber identification number is extracted from the request message transmitted from the transcribe client device, and the extracted identification number and the transcriber management table (see FIGS. 32 and 33) are matched and given priority. Extract the work contents to be transcribed, and transcribe the extracted work contents and search all data distribution files (refer to the process status of the middle process of the data manager for the explainer or the customer in order, not allocated or partially allocated Search against completed). If there is data that matches the work content to be prioritized, the process proceeds to 2240. In other cases, it is checked whether there is any transcription data to be further processed. If 2238, otherwise 2
The processing is shifted to 239 (2235-2237).

  When instructed to shift to 2238 in 2237, the fact that there is no transcription data in response to the transcription request is returned to the requesting client, and this processing is terminated (2238).

  In addition, when the transition to 2239 is instructed in 2237, data is extracted from the transcription data distribution file in the oldest order and the work is scheduled to be allocated (2239).

  Next, in order to check whether or not there is a duplicate assignment for the scheduled transcription data, refer to the corresponding middle process worker ID section, check the transcript identification number, and if they match, prevent duplicate assignment. Therefore, if another next data is searched and there is a match, the process proceeds to 2235, and if it does not match, the process proceeds to 2242 (2240, 2241).

  If there is a discrepancy in 2241, determine the range of transcription data for the transcription request, and embed a 4 digit transcriber identification number (eg A001) in the employee ID area in the data manager for the explainer or customer. Then, the processing status is changed to “1” (partially assigned) or “2” (all assigned) (2242).

Next, the data of the determined transcription range is returned to the requesting client, and this processing is terminated.
An example of the range of transcription is described, for example, when the transcripter identification number A001 makes a transcription request, the data range in the bold line frame in FIG. 32 becomes the target of transcription.

  FIG. 27 shows an example of a transcription work distribution table, which includes a similarity degree range in the middle process, a number of transcription divisions, a transcription aggregation number, a similarity degree range after division / aggregation, and the number of transcription workers.

  FIG. 28 shows the data contents of the transcription data management unit for distribution.

  FIG. 30 is an example illustrating the contents of organization during the division and aggregation of the middle steps of the data management unit.

  Also. FIG. 32 shows an example of assigning a small process identification number to the assumed explanation definition body at the time of the middle process reorganization. It is also an image of the definition of the middle process and the small process of the assumed explanation definition body for the transcription data distribution.

  Further, FIG. 33 is a table for managing the transcribers. The number of persons to be managed, the transcriber identification number, the skill level, the assignment priority work (1), the assignment priority work (2), and the assignment priority work (3). Consists of.

Next, FIG. 34 shows a setting example (setting example after the number of management targets) when there are 10 transcribers.
A person with a low skill level works by giving priority to a similarity of 81 or more, centering on the transcript of the presenter. The skill level intermediate person works by giving priority to a similarity degree of 31 or more, centering on the transcript of the customer.

  Next, processing that assumes a case where a person with a high skill level works by giving priority to a similarity degree of 30 or less will be described.

  FIG. 34 is a flowchart of the processing control unit 31 of the transcription client, and will be described according to the processing order.

First, after selecting the transcription work start on the transcription initial screen (FIG. 36), the transcription work deadline included in the received transcription data is notified to the transcription person on a pop-up screen. (3401)
Next, after confirming the pop-up screen contents, press the confirmation key to delete the contents. After that, the transaction for starting the transcription work is selected and the work is started. (3402)
Then, edit the received transcription data for screen display, display the transcription work screen, and wait for various event occurrences to perform the corresponding processing each time an event such as pressing various instruction buttons or input data occurs. I do. (3403, 3404)
(FIG. 36 shows an example of the entire transcription work screen image, and FIG. 37 shows a specific screen display example of the transcription work.)
Here, various event occurrences are waited, and when the event occurs, the event content is checked (3405). If it is determined that the utterance icon is clicked, the process proceeds to 3406. If the sound reproduction system button is clicked, the process proceeds to 3407.

  If it is determined in 3405 to shift to 3406, the audio data logically coupled to the small process unit chorded to the corresponding icon is reproduced (3406). Then, the process proceeds to 3404.

  If it is determined in 3405 to shift to 3407, it is further confirmed whether the button is an audio reproduction system button. If so, the process shifts to 3408, and if not, the process shifts to 3409 (3407).

  If it is determined in 3406 that the audio playback system button has been pressed, for example, if the playback button is pressed, the start time associated with the time designation bar is obtained, and the continuous audio data is played back at the corresponding start time. The time bar is moved according to the elapsed time by sound reproduction. When the other button is pressed, the corresponding function operation is executed. (3408) Thereafter, the process proceeds to 3404 to wait for various event occurrences.

  Further, in 3406, if it is not determined that the audio playback system button has been pressed, the presence / absence of a click on the window scroll is checked. If yes, the process moves to 3410, and if not, the process moves to 3411. (3409).

  If it is determined that the process proceeds to 3410, the corresponding window scroll is performed by clicking the corresponding window scroll button (3410). Thereafter, the processing shifts to 3404 in order to wait for various event occurrences.

  If it is determined in 3409 to shift the process to 3411, it is determined whether or not the keyboard is pressed. If so, the process shifts to 3413, and if not, the process shifts to 3412.

Here, if it is determined that the processing is shifted to 3412, it is other processing, processing corresponding to that is performed, and the processing is shifted to 3404. (3412)
If it is determined that the process is to be moved to 3413, it is further checked whether or not the transcription end key is pressed. If it is the transcription end key, the process is transferred to 3415. The processing proceeds to 3414 assuming that it is an input (3413).

Here, when the processing is shifted to 3414, it is regarded as a transcription data input,
Save the transcription data. Thereafter, the processing shifts to 3404 in order to wait for various event occurrences. (3414)
When the process is shifted to 3415, the transcription data and the like are saved in a file by a transcription termination operation, the process returns to the initial screen, and the process is terminated.

  FIG. 36 shows an example of the initial screen for transcription. FIG. 37 is an overall image diagram of the transcription work screen. The outline of each item is as specified in FIG. For example, when the transcriber identification number A001 makes a transcription request, a specific screen display example of the transcription operation displayed is shown in FIG. In particular, (5) “Progress process” to (8) “Unexpected keyword” information displays in the figure can help you understand the level of difficulty of the transcription data and the time required for the work, and eliminate concerns about strict adherence to the submission deadline. Can be expected.

  FIG. 39 is a flowchart of the transcription data collection means 105, which will be described in the order of processing.

  First, the received transcription data is referred to and it is checked whether or not the transcription collection data based on the corresponding transcription distribution data has already been generated (3801).

  If the result of the check is that the file has been generated, go to 3804. In other cases, the processing shifts to 3803 (3802).

In 3803, transcription collection data is collectively generated based on the corresponding transcription distribution data, and the process proceeds to 3804. Here, the folder and file structure of the generated data is as shown in FIG. 40, which is the same structure as the transcription distribution file shown in FIG. 29, and the data is also generated by taking over the contents of the distribution file. (3803)
In 3804, the received transcription data is collected, and the input data for the transcription is accumulated in the transcription data collection file. (3804) Specifically, the character string data input in the “transcription data” item is additionally stored in the “transcription input data” item of the corresponding explainer or customer in the corresponding definition body. At the time of additional storage, store so that the number of cases and input data can be understood.
Further, the location where the transcriber identification number recorded in the “middle process worker ID” in the data management unit of the corresponding instructor or customer is matched with the current operator identification number is obtained, and the corresponding control flag 1 digit (initial (Value = 0) is changed to “1” (collected for the corresponding identifier). Further, the processing status is changed to “1” (partially collected) or “2” (all collected). (See Figure 42)
FIG. 41 shows an example of a transcription data collection file.
FIG. 42 is an overview of the data management unit of the transcript data collection file of the explainer or customer, and is the extent to which one digit of the control flag is added to the middle process worker ID.

  FIG. 43 is a flowchart of the transcription edit data creation means 106, which will be described in the order of processing.

  First, the data managers for both the explainer and the customer corresponding to the collected transcription data are checked to check whether or not all the intermediate process statuses are “2” (all collected) (4201). . If all the check results are “2”, go to 4203. In other cases, the present process is terminated (4302).

Corresponding edit data presentation files are generated in batch for the transcription editor. The folder and file structure of the generated data are shown in FIG. 44, and the data management unit is integrated into one. Data is also transcribed and the contents of the data collection file are taken over and generated. FIG. 45 shows an overview of the data management unit of the transcription edit data file, and FIG. 46 shows an example of the intermediate process and the small process of the assumed definition body.

  Note that the reorganization progress small process that existed at the time of the transcription distribution file is unnecessary and does not exist.

  FIG. 47 is a flowchart of the processing control unit 107 of the editing client, which will be described according to the processing order.

  First, after selecting the start of editing work on the editing initial screen (FIG. 49), the editor is notified of the editing work deadline included in the received editing data on a pop-up screen (4601).

  Next, after confirming the pop-up screen contents, press the confirmation key to delete the contents. Thereafter, the transaction for starting the editing work is selected and the work is started (4602).

  Here, the pointer in the assumed explanation definition body of the received editing data is set to the first progress small process identification number position (4603).

Further, referring to the “transcription input data” item for the explainer pointed to by the pointer, all the input data is displayed in the “draft transcription data” item on the editing work screen, so that all are copied. Also, the input data of the “transcription input data” item are compared with each other, and if there is a mismatch, editing is performed so that “draft data mismatch” is displayed in red in the “mismatch presence / absence” item on the editing work screen.
If they match, “none” is edited (4604).

At this time, all the input data is displayed in the “draft transcription data” item in the editing work screen with reference to the customer “calling input data” item pointed to by the pointer. Also, the input data of the “transcription input data” item are compared with each other, and if there is a mismatch, editing is performed so that “draft data mismatch” is displayed in red in the “mismatch presence / absence” item on the editing work screen.
If there is a match, edit “none”.

  Thereafter, it is checked whether the pointer pointed to by the assumed explanation definition body is final. If the pointer is final, the process proceeds to FIG. 48: 4609; otherwise, the process proceeds to 4608 (4607).

  When the processing is shifted to 4608, the pointer of the assumed explanation definition body is updated to the next progress small process identification number position, and the processing shifts to 4604 (4608).

  48: When the processing is shifted to 4609, other received data is also edited for screen display, an editing work screen is displayed, and each time an event such as pressing of various instruction buttons or input data occurs. In order to execute the corresponding process, various event occurrence waits are performed (4609). Incidentally, FIG. 55 shows an example of the entire image of the editing work screen.

After that, it waits for various events to occur and checks the event contents when an event occurs. (4610)
First, when the utterance icon is clicked, the audio data logically coupled to the small process units linked to the corresponding icon is reproduced. Thereafter, the processing shifts to 4610 to wait for various event occurrences (4611, 4612).

Also, in the case of clicking on the audio playback system button, for example, when the playback button is pressed, the start time tied to the time designation bar is obtained, and the continuous audio data is played back at the corresponding start time. The time bar is moved according to the elapsed time by sound reproduction. When the other button is pressed, the corresponding function operation is executed. Thereafter, the flow goes to 4609 to wait for various events. The processing is shifted to (4613, 4614).

  Further, when the window scroll is clicked, the corresponding window scroll is performed by clicking the corresponding window scroll button. Thereafter, the processing shifts to 4609 to wait for various event occurrences (4615, 4616).

  In the case of a keyboard system such as a key press (4617), it is further checked whether or not the transcription end key is pressed (4619). Save, return to the initial screen, and end this process (4621).

Others are regarded as transcription data input and data collection data input, and the edited data is saved. At this time, if “draft data mismatch” is edited in the “mismatch presence / absence” item in the editing work screen, it is re-edited to “none”, and the process proceeds to 4609 (4620).
In the case of other events, other processing is performed. Thereafter, waiting for various events to occur. FIG. 51 is an example of an image of a transcription completion data file.

  FIG. 52 is an outline of the data management unit of the transcription completion data file.

The block diagram of the question answering system in one embodiment of this invention Block diagram of a question answering system in an embodiment of the present invention The flowchart in one embodiment of this invention The flowchart in one embodiment of this invention The flowchart in one embodiment of this invention The flowchart in one embodiment of this invention The flowchart in one embodiment of this invention The figure explaining the outline | summary of the sentence label process in one embodiment of this invention Explanatory drawing of the table utilized by the sentence label identification process in one embodiment of this invention Explanatory drawing of the table utilized by the sentence label identification process in one embodiment of this invention Explanatory drawing of the table utilized by the sentence label identification process in one embodiment of this invention The figure which showed the process of the sentence label identification process in one embodiment of this invention The figure which showed the process of the sentence label identification process in one embodiment of this invention The flowchart in one embodiment of this invention The flowchart in one embodiment of this invention The flowchart in one embodiment of this invention The flowchart in one embodiment of this invention Example of screen display on the side of the presenter in one embodiment of the present invention Example of screen display on the side of the presenter in one embodiment of the present invention Explanatory drawing of the recording data of the presenter's actual voice start time, voice elapsed time, utterance keyword, unexpected keyword, similarity obtained by the process control / communication means acquired in one embodiment of the present invention Explanatory drawing of the description content collection file in one embodiment of this invention. Explanatory drawing of the data management part in one embodiment of this invention. The flowchart in one embodiment of this invention. The flowchart in one embodiment of this invention. The flowchart in one embodiment of this invention. The flowchart in one embodiment of this invention. The figure which showed the example of the data for transcription distribution in one embodiment of this invention. The figure which showed the example of the data for transcription distribution in one embodiment of this invention, The figure which showed the data recorded on the data management part in one embodiment of this invention. The figure which showed the mode of the reorganization (division | segmentation) of the middle process in the data management part in one embodiment of this invention. The figure which showed the data recorded on the data management part in one embodiment of this invention in one embodiment of this invention. The figure which showed the transcription management table in one embodiment of this invention. The figure which showed the transcription management table in one embodiment of this invention. The figure which showed the example of a setting (setting example after the number of management object persons) in case there are ten transcripts in one embodiment of this invention. The flowchart in one embodiment of this invention. The figure which showed the transcription initial screen in one embodiment of this invention. The figure which showed the screen display of the transcription work in one embodiment of this invention. The figure which showed the screen display of the transcription work in one embodiment of this invention. The flowchart in one embodiment of this invention. The figure which showed the data structure at the time of the data collection for transcription collection in one embodiment of this invention. The figure which showed the example of the image of the transcription data collection file in one embodiment of this invention. The figure which showed the outline | summary of the data management part of the transcription data collection file of the explainer or customer in one embodiment of this invention The flowchart in one embodiment of this invention. The figure which showed the outline | summary of the data management part of the transcription data collection file in one embodiment of this invention The figure which showed the outline | summary of the data management part of the transcription edit data file in one embodiment of this invention The figure which showed the image example of the middle process of the assumption definition body in one embodiment of this invention, and a small process. The flowchart in one embodiment of this invention. The flowchart in one embodiment of this invention. The figure which showed the edit initial screen in one embodiment of this invention. The figure which showed the example of the whole image of the edit operation | work screen in one embodiment of this invention. The figure which showed the image example of the transcription completion data file in one embodiment of this invention. The figure which showed the outline | summary of the data management part of the transcription completion data file in one embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Product description characterization assistance apparatus 21 Client apparatus for branch offices 31 Client apparatus for transcripts 41 Client apparatus for editors 51 Client apparatus for compliance personnel

Claims (7)

On the computer to support the request of work that documented the presenter and Kai story of the customer to perform a description of the goods,
Calculated results and similar degree of the description contents according to the level and the product of the customer defined definition of the speech recognition for the speech in accordance with the conversation,
According to the degree of similarity , the voice is divided into a plurality of transcription ranges,
According to the number of the transcription range, a product description content character, which causes a terminal operated by each of a plurality of transcribers to perform the work to transmit data related to the transcription range. Support program. In addition to the computer,
Wherein if the similarity degree is equal to or larger than the predetermined, the text of the description contents defined previously Kijo prosthetic body, Product Description of claim 1, characterized in that to execute a process of outputting as an input candidate of the documentation Content characterization support program. Support child computer the request of work to document the meeting talking about the description and the customer to perform a description of the products,
Calculated results and similar degree of the description contents according to the level and the product of the customer defined definition of the speech recognition for the speech in accordance with the conversation,
According to the degree of similarity , the voice is divided into a plurality of transcription ranges,
A product description content character that executes a process of transmitting data related to the transcription range to a terminal operated by each of a plurality of transcripters performing the work according to the number of the transcription range Support method. The computer further includes:
Wherein if the similarity degree is equal to or larger than the predetermined, the text of the description contents defined previously Kijo prosthetic body, Product Description of claim 3, characterized in that executes processing for outputting as an input candidate of the documentation Content characterization support method. A product description content lexicalization support device that supports a request for an operation to document a conversation between a customer who explains a product and a customer,
The results of the speech recognition for the speech according to the conversation with obtaining the degree of similarity between the description contents according to the level and the product of the customer defined in definition body, depending on the degree of similarity, a plurality of the voice a data distribution means for dividing the transcription in the range,
Depending on the number of the transcription range, contents of description of item in which a plurality of transcription who performs the work to a terminal to be operated, respectively, and having a communication means for transmitting data relating to the transcription range Character support device . Said communication means, when the similarity degree is equal to or larger than a predetermined, product description of claim 5, the text of the description contents defined in the definition, and outputs as an input candidate of the documentation Content characterization support device. On the computer,
According to the similarity between the recognition result of the input sound and the content defined in the definition body, the input sound is divided into a plurality of pieces to obtain a plurality of divided sounds.
A product description content characterizing support program for causing a plurality of terminals capable of reproducing and inputting characters to the plurality of divided voices to execute a process of distributing and transmitting the plurality of divided voices.