JPH1125112A - Method and device for processing interactive voice, and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an interactive voice processor improved so as to efficiently reproduce the information of a necessary part from recorded interactive voices. SOLUTION: A division processing part 1 processes interactive voices inputted in time series order to generate plural voice data. An item contents recording processing part 2 allows contents to correspond to a data item (topic) to be an index of interactive voices. A data item correspondence processing part 3 allows the data item to correspond to a voice data group and stores a voice data set in a data storage part 4 in each data item. An interactive contents reproducing part 6 retrieves only a voice data set for a necessary data item from the data storage part 4 and reproduces the contents of an interaction.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice data processing technique for recording a dialogue voice, for example, and for efficiently reproducing the dialogue voice afterwards to grasp the content of the dialogue.

[0002]

2. Description of the Related Art In a service system to which an interactive voice processing technology is applied, a support apparatus for a receptionist who creates a customer correspondence record while interacting with a customer is known. In this support device, the receptionist inputs customer data items determined in advance using a keyboard or selects a menu, thereby inputting customer data and creating a customer response record. The service technician sees the customer's request written in the customer affairs record created by the reception clerk, determines the content of the response to the customer, and takes necessary measures.

[0003]

[0005] By the way, if the information transmitted from the receptionist is insufficient to deal with the customer, the service representative needs to contact the customer again. Occurs. In this case, for the customer,
Since the same contents as those already spoken to the receptionist have to be told again to the service person, it is troublesome. The solution to this problem is to record all interactions between the customer and the receptionist so that service personnel can understand the details of the conversation ex post if insufficient information is provided. It is possible. However, in this method, a part of the recorded dialogue that is actually required is very short, but the entire dialogue must be heard. Therefore, when the conversation is long or when a plurality of conversations are recorded, there is a problem that it takes time for the service staff to hear the conversation voice.

[0004] Such a problem is caused not only by a service technician afterhand ascertaining the contents of the dialogue between the customer and the receptionist, but also by listening to the recorded dialogue voice for the purpose of simply knowing the outline of the dialogue. This also occurs in the case.

SUMMARY OF THE INVENTION It is an object of the present invention to provide an improved dialogue speech processing method which can efficiently reproduce necessary information from recorded dialogue speech. Another object of the present invention is to provide an apparatus suitable for implementing the above-described interactive voice processing method, and a recording medium for realizing the apparatus on a general-purpose computer device.

[0006]

A dialogue speech processing method according to the present invention for solving the above-mentioned problem is a method using a computer device, which divides an inputted dialogue speech into a plurality of speech data and divides the speech data into a plurality of speech data. The method includes a step of storing each voice data in association with each topic, selectively reading out a voice data group corresponding to a specific topic from the stored voice data, and reproducing a conversation content. According to this method, only the voice data group on the topic can be searched for and played back, so that the search efficiency is improved and the time required for grasping the conversation contents is shorter than when all voice data is played back in chronological order. It can be significantly shortened.

[0007] Further, another conversational speech processing method according to the present invention comprises:
The input dialogue voice is divided into a plurality of voice data, and it is determined for each voice data whether a keyword determined in advance for each topic is included, and the voice data is associated with the determination result. Storing the voice data including the keyword from the stored voice data in the storage order and reproducing the dialogue content. According to this method, the keyword is included,
That is, the contents of the dialogue are reproduced in chronological order based on the voice data whose importance is estimated to be relatively high, so that it is possible to grasp the outline of the contents of the dialogue.

It is also possible to assign priorities to the keywords in advance, and to reproduce the audio data corresponding to the keywords having the higher priority.

According to a first aspect of the present invention, there is provided a dialogue speech processing apparatus for solving the above-mentioned problems, comprising: a speech input unit for inputting a dialogue voice; Means for generating data, and means for accumulating the plurality of generated voice data in association with data items each serving as an index of the content of a conversation, and for each of the data items from the plurality of stored voice data. It is characterized in that it is configured to search for a corresponding audio data group.

[0010] The conversational speech processing apparatus of the second configuration comprises a speech inputting means for inputting a conversational speech, a means for dividing the inputted conversational speech for each predetermined processing unit and generating a plurality of speech data, Means for determining whether a keyword determined for each voice exists in each voice data, and means for adding the determination result information to the voice data and storing the same, and from the plurality of stored voice data, The audio data group including the keyword is searched in the order of accumulation.

According to a third aspect of the present invention, there is provided a dialogue speech processing apparatus comprising: a voice input unit for inputting a dialogue voice; a unit for dividing the input dialogue voice into predetermined processing units to generate a plurality of voice data; Means for determining whether a keyword determined for each voice is present in each voice data, and adding information on the result of the determination to each voice data, and converting each voice data into a data item serving as an index of a conversation content. Means for associating and storing a plurality of voice data, and searching for a corresponding voice data group for each data item or each keyword from the plurality of stored voice data.

[0012] Each of the dialogue voice processing devices preferably further comprises a reproducing means for reproducing the content of the dialogue based on the searched corresponding voice data group.

Further, a recording medium of the present invention for solving the above-mentioned other problems is a recording medium in which a program for causing a computer device to execute the following processing is recorded in a form readable by the computer device. (1) a voice input process for inputting a dialogue voice, (2) a process for generating a plurality of voice data by dividing the input dialogue voice for each predetermined processing unit, and (3) a keyword previously determined for each topic. A process of determining whether or not the voice data is present in each voice data; (4) assigning information on the result of the determination to the generated voice data, and associating each voice data with a data item serving as an index of the dialogue content (5) a process of retrieving the stored voice data for each of the data items or for each keyword and reproducing the contents of the dialogue.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an embodiment in which the present invention is applied to a dialogue voice processing apparatus for a dialogue voice exchanged between a customer and a receptionist will be described in detail.
FIG. 1 is a schematic configuration diagram of a conversational speech processing device according to the present embodiment. This interactive voice processing device is formed by a computer device reading and executing a predetermined program, which is formed by a division processing unit 1, an item content recording processing unit 2, a data item association processing unit 3, a data storage unit 4, a keyword It has the functions of the detection unit 5 and the dialogue content reproduction unit 6. A program for forming each of the functional blocks 1 to 6 is normally stored in an internal storage device or an external storage device of the computer device, and is read and executed as needed. May be stored in a recording medium that can be separated, for example, a CD-ROM, an FD, or the like, and may be installed in the internal storage device or the external storage device at the time of use and provided for execution at any time.

The dialogue voice according to the present embodiment is a voice of a customer and a receptionist input to the apparatus through a telephone or the like. The division processing unit 1 divides the input dialogue voice into voice data of an appropriate processing unit, and assigns a continuous number to each voice data. The division of the audio data can be performed, for example, using a silent section having an arbitrary length or more between utterances. In many cases, the dialogue voice is divided as one section of voice data from when one speaker starts talking to when the other speaker starts talking. Of course, it is also possible to perform the division processing by another method.

The item content recording processing unit 2 extracts the contents of a predetermined customer correspondence data item (hereinafter, abbreviated as a data item) from the input dialogue voice, and outputs the extracted result to the data item association processing unit 3. hand over. The extraction of the item contents from the conversation voice can be performed by voice automatic recognition processing or the like. For example, the receptionist who operates the apparatus asks the customer for the data item, and performs voice recognition processing on the voice of the reply portion of the customer to grasp the contents of the data item for the customer. Alternatively, when the receptionist talks directly with the customer without using the automatic voice recognition processing and the like and the contents related to the data items are included in the conversation, the data is sent to the item contents recording processing unit 2 by a keyboard (not shown). You may make it input as item content.

The data item association processing unit 3 associates a data item serving as an index of a topic between a customer and a receptionist with each item content extracted by the item content recording processing unit 2, and performs a division process. The audio data (number assigned) from the unit 1 is associated with the data items, and the data items are stored in the data storage unit 4 as a series of customer correspondence data. In addition, for each data item, a key input by the receptionist (keyboard input, the same applies hereinafter)
All of a series of voice data spoken before the time point when there is a key input of the immediately preceding data item and after the time point when the key input of the previous data item was input are associated with the data item.

The keyword detection unit 5 holds in advance a list of keywords expected to appear in a dialog for each topic, that is, for each data item, and determines whether the above-mentioned keywords are included in the divided voice data. The data is checked by the recognition process, and information on the presence or absence is stored in the data storage unit 4 in association with the voice data. This keyword recognition process is performed in parallel with the voice data division process.
Execute in multitasking format.

The conversation content reproducing unit 6 retrieves the data corresponding to each data item or each keyword from a plurality of voice data stored in the data storage unit 4, arranges them in a predetermined order, and reproduces the conversation content. . Upon playback,
Normally, an acoustic technique is used. However, a method of performing speech recognition processing on speech data, converting the speech data into character information, encoding, and graphics, and displaying the resulting data on a display may be employed. A search method such as a text search can be employed for searching for audio data.

Next, a process for storing voice data in the dialogue voice processing apparatus of the present embodiment (hereinafter, this apparatus) will be described with reference to FIG. It is assumed that the storage processing of the voice data in the present apparatus continues until the customer connects and disconnects the telephone line.

In this apparatus, after confirming that the line has not been disconnected (step S1: No), the apparatus waits for key input and voice input (step S2: No, step S3: N).
o). If a voice is input (step S3: Ye
s) A recording process for each audio data in a recording area (not shown) is performed (step S4). This process is repeated until the voice input ends (step S5: No). When the voice input ends (step S5: Yes), a numbering process is performed on a series of voice data divided from the input voice (step S6). The above processing is performed by the division processing unit 1. In step S2, when there is a key input, for example, two voice data are used for a certain data item, and when there is a key input for that data item, the data item association processing unit 3 A process of associating with audio data is performed (step S7). Thereafter, the keyword detection unit 5 detects the presence or absence of the keyword in the keyword recognition processing and the voice data,
The detection result is stored in the data storage unit 4 (step S
8). These series of processes are repeated until the line is disconnected.

A specific description will be given of a case where the above-described processing is applied to a case where a customer provides a consultation regarding a purchased personal computer to a receptionist of the present apparatus. FIG. 3 is a diagram showing the flow of the dialogue in this example and the contents of the voice data at that time. In this example, a dialogue exchanged between a customer and a receptionist is divided into a plurality of audio data in the division processing unit 1, and serial numbers "0000" to "0013" are assigned to each audio data. Also, a voice data group corresponding to the data item, that is, the voice data set start voice data “0000” is recorded in the corresponding area of the data storage unit 4. Dialogue progresses,
Once the customer's name is known, the receptionist types the customer name "Isobe" into the data field "Customer Name"
These audio data sets (0001 to 0002) are stored in corresponding areas of the data storage unit 4. Similarly, the data item “message” is “memory model number inquiry”, and the data item “target hardware” is a computer type “XX-Y”.
Y ", the customer telephone number" 03-
2456-7777 "by key input, and a voice data set for each data item is stored in the corresponding area of the data storage unit 4.

FIG. 4 shows an example of data (customer correspondence data) representing the correspondence between data items and their contents created based on this dialogue. In this embodiment, when the key input which is the delimiter information for the data item is completed, the most recently uttered voice data before that is used as the end voice data of the voice data set for the data item. As a result, the range of the audio data set and the corresponding item are determined.

The keyword detecting unit 5 determines when the range of the voice data set and the corresponding item are determined, for example, when the data item “message” is key-inputted to the content “memory model number inquiry”. Then, it is determined that the voice data sets “0003” to “0005” are voice data corresponding to the data item “message”, and the keyword recognition processing is performed on the voice data “0003” to “0005”. . This keyword recognition process is performed based on a list of keywords registered in advance. FIG.
Is a diagram showing an example of this list. For example, "PC", "I want to listen", "Model number", "Tell me", etc. are registered as keywords corresponding to the data item "Message".・ There is. For example, the keyword detection unit 5 outputs the voice data “0003” and “000” as voice data including a keyword for the data item “message”.
5 "is selected, and it is determined that the voice data" 0004 "does not include a keyword. As shown in FIG. 6, these results are, as shown in FIG. 6, data item names, item contents, and corresponding audio data (one or more audio data numbers, audio data file names, 1 = presence / 0 = absence) are stored in the data storage unit 4 as set data.

By the way, in the example of FIG. 3, information on four data items is inputted by a key by the receptionist. However, among the contents spoken by the customer, regarding the information of “purchase time last September”, the purchase place “XX Denki”, and the memory capacity to be purchased “change from 8M to 40M”, The receptionist has not entered a key.
Also, the customer's telephone number has been incorrectly keyed in. In such a case, if the service rep attempts to reply based on the customer response data created by the reception rep, it is expected that the information will be insufficient or the information to be answered will be incorrect. In this case, the service person will respond as follows.

First, when there is a fact that the model number of the memory differs depending on the production time, it is necessary to correctly know the product number and the purchase time. At this time, the voice data set of the "message" part which has already been recorded is retrieved from the data storage unit 4 and played back by the conversation content playback unit 6 to know the purchase time of the product. And the correct memory model number. In addition, when a call is made to the telephone number of the "contact" item of the customer correspondence data, the receptionist does not connect to the customer due to an incorrect key entry. In this case, the voice data set corresponding to the “contact information” item of the customer correspondence data is selectively retrieved from the data storage unit 4 and played back by the conversation content reproduction unit 6 to be confirmed. You will be able to know the correct phone number.

Furthermore, when listening to the outline of the contents of the dialogue,
By retrieving only the voice data containing the keyword from the data storage unit 4 in the order of time and sorting the data, and reproducing this by the conversation content reproduction unit 6, the outline can be quickly grasped. Become like FIG. 7 shows FIG.
21 is an example of the contents of a dialog created by extracting only voice data including a keyword from the example of the dialog. The underlined words are keywords registered in advance. As is clear from FIG. 7, the content is much shorter than the content of the dialog shown in FIG. In addition, since all necessary conversation contents are covered, correct understanding of the conversation contents is possible.

When the contents of the dialogue are reproduced, the audio data is reproduced after being sorted, or the keywords are prioritized according to an arbitrary rule, and the audio data corresponding to the keywords having the higher priority are reproduced in order. It is also possible to do so. In this way, the conversation voice can be grasped in the order of interest, and a use form that matches the purpose can be realized.

As described above, in the dialogue speech processing apparatus of the present embodiment, the dialogue voice exchanged between the customer and the receptionist is divided to generate a plurality of voice data, and is determined in advance. Since data is stored in association with each audio data set for each data item, and playback can be performed for each data item, when grasping the contents of the dialogue ex post facto, the sound is passed through from the beginning of the dialogue as in the past. There is no need to listen, and the labor and time required to obtain the necessary information can be saved.

Also, of the divided voice data, important ones concerning the contents of the dialogue, for example, only the voice data including a keyword predetermined for each topic are arranged in a time series so that the outline contents of the dialogue can be reproduced. Therefore, it is possible to easily understand the contents of the dialogue.

Although the present embodiment has been described by taking as an example the dialogue voice between the customer and the receptionist, the present invention is not limited to this example, and the present invention can be applied to any dialogue. It goes without saying that there is something.

[0032]

As is apparent from the above description, according to the present invention, when recording a dialogue voice, the dialogue voice is divided into a plurality of voice data and stored as a voice data set for each topic. Therefore, it is possible to reproduce only a voice data set of a topic portion that may include information about a topic to be known.

Further, according to the present invention, voice data including a required keyword can be selectively searched out from a plurality of stored voice data and reproduced, so that the contents of the dialog can be summarized as a whole. It becomes possible to grasp. In particular, by reproducing the voice data including the keyword in the chronological order of the dialogue, it is possible to reproduce the voice along with the flow of the original dialogue voice, so that the contents of the dialogue can be correctly understood.

[Brief description of the drawings]

FIG. 1 is a schematic functional explanatory diagram of a conversational voice processing device according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a processing procedure of the conversational voice processing device of the embodiment.

FIG. 3 is an explanatory diagram showing a flow of a dialogue between a customer and a receptionist of a dialogue voice processing apparatus and the contents of voice data at that time.

FIG. 4 is an explanatory diagram of customer correspondence data according to the embodiment.

FIG. 5 is an explanatory diagram showing an example of the content of a preset keyword list.

FIG. 6 is an explanatory diagram showing an example of the contents of group data stored in a data storage unit.

FIG. 7 is an explanatory diagram of a dialogue voice reproduced by keyword extraction.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Division processing part 2 Item content recording processing part 3 Data item correspondence processing part 4 Data storage part 5 Keyword detection part 6 Dialogue content reproduction part

Claims (9)

[Claims] An input dialogue voice is divided into a plurality of voice data, the divided voice data is stored as a voice data set for each topic, and voice data corresponding to a specific topic is stored from the stored voice data. A dialogue speech processing method using a computer device, comprising a step of selectively retrieving a set and playing back the content of the dialogue. 2. The range of the audio data set is such that the audio data input after the previous determination of the delimitation information for each topic is the start audio data, and the latest input audio data before the determination of the next delimitation information is set. Is determined as the end voice data. 3. An input dialogue voice is divided into a plurality of voice data, and it is determined for each voice data whether a keyword predetermined for each topic is included, and the voice data is determined. A dialogue voice processing method using a computer device, the method including a step of storing voice data including the keyword from the stored voice data in the storage order and playing back the content of the dialogue while storing the voice data in association with a result. 4. The interactive voice processing method according to claim 3, wherein priorities are assigned to the keywords in advance, and voice data corresponding to the keywords having a higher priority is reproduced preferentially. 5. A voice input unit for inputting a dialogue voice, a unit for generating a plurality of voice data by dividing the input dialogue voice for each predetermined processing unit, Means for storing in association with a data item serving as an index of a dialogue content, and a dialogue voice processing apparatus configured to search for a corresponding voice data group for each data item from the stored plurality of voice data. . 6. A voice input unit for inputting a dialogue voice, a unit for generating a plurality of voice data by dividing the input dialogue voice for each predetermined processing unit, and Means for determining whether or not the voice data exists in the voice data; and means for adding the determination result information to the voice data and storing the voice data. The voice data group including the keyword is stored from the plurality of stored voice data. An interactive speech processing device configured to search in the order of accumulation. 7. A voice input unit for inputting a dialogue voice, a unit for generating a plurality of voice data by dividing the input dialogue voice for each predetermined processing unit, and Means for determining whether or not the voice data is present; and means for adding information on the result of the determination to each voice data and storing each voice data in association with a data item serving as an index of a dialogue content. An interactive voice processing device configured to search for a corresponding voice data group for each data item or each keyword from the plurality of stored voice data. 8. The dialogue voice processing apparatus according to claim 5, further comprising a reproducing unit for reproducing the content of the dialogue based on the searched corresponding voice data group. 9. A voice input process for inputting a dialogue voice, a process for dividing the input dialogue voice for each predetermined processing unit to generate a plurality of voice data, A process of determining whether or not the voice data exists in the voice data; a process of adding the determination result information to the generated voice data, and accumulating each voice data in association with a data item serving as an index of a conversation content. A recording medium in which a program for causing a computer device to execute the process of retrieving the stored voice data for each data item or each keyword and reproducing the contents of a conversation is recorded in a form readable by the computer device.