JPH10322450A - Voice recognition system, call center system, voice recognition method and record medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To relieve the load of the operator by enhancing the recognition accuracy. SOLUTION: A main control section 15 receives a voice of the user and supplies it to the operator via a speaker of an operator operation section 12 and instructs voice recognition to a voice recognition section 13. The voice recognition section 13 recognizes the user voice and gives the result of recognition and its recognition rate to the main control section 15. The main control section 15 displays the recognition result on a display device of the operator operation section 12. The operator recites the voice of the user while referring to the displayed recognition result. The main control section 15 instructs the recognition of the voice recited by the operator to the voice recognition section 13. The voice recognition section 13 recognizes the voice of the operator and gives the result of recognition and its recognition rate to the main control section 15. The main control section 13 selects either of the results of recognition received from the voice recognition section 13 which has a higher recognition rate as a final recognition result.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an interactive speech recognition system for recognizing a user's speech, and more particularly to a speech recognition system and a recording medium capable of improving recognition accuracy. Speech recognition systems, call center systems,
The present invention relates to a voice recognition method and a recording medium.

[0002]

2. Description of the Related Art As a system for receiving inquiries, orders, and the like from a user by telephone, for example, an automatic response type voice recognition system without an operator is used.
In such a system, the computer recognizes the voice of the user from the telephone and automatically performs processing based on the recognition result. Further, there has been proposed a voice recognition system in which an operator responds to a telephone call from a user, and a computer recognizes a voice read by the operator and performs input processing.

[0003]

However, in the case of the automatic answering type system, since the voice from the telephone is recognized, the recognition accuracy is low, and the user has to repeat the utterance many times until it is correctly recognized. . Further, in the case of a system that recognizes voices repeated by an operator, if the user utters correctly, but the operator erroneously repeats, it is necessary to repeat the utterance.

[0004] The present invention has been made in view of the above situation, and has as its object to provide a speech recognition system, a call center system, a speech recognition method, and a recording medium with high recognition accuracy. Speech recognition systems, call center systems,
Another object is to provide a voice recognition method and a recording medium.

[0005]

In order to achieve the above object, a speech recognition system according to a first aspect of the present invention comprises:
In a voice recognition system that receives a message from a user by voice, a receiving unit that receives the voice of the user, and analyzes the voice of the user that is received by the receiving unit. A first recognition unit that generates a first recognition probability for a first recognition result, a supply unit that supplies a voice of the user to a manager, an input unit that inputs a voice of the manager, A second recognition unit that analyzes a voice of the manager input by the input unit and generates a second recognition result and a second recognition probability for the second recognition result; and the first and second recognition units. Determining means for determining whether or not the first and second recognition results generated by the recognition means match; and determining that the first and second recognition results match, determining whether the first and second recognition results match. Determine the second recognition result as the final recognition result And determining means, when the first and second recognition result is judged not to match, the first and second
Means for comparing the recognition probabilities of the two and selecting the result as the final recognition result.

According to such a configuration, recognition processing is performed on both the user's voice and the manager's voice, and if the recognition results are different, the two recognition probabilities are compared to determine the appropriate recognition result. Is adopted. Thereby, a speech recognition system with higher recognition accuracy can be realized.

[0007] When the first recognition probability is larger than the second recognition probability, the selection means selects the first recognition result as a final recognition result, and the first recognition result is determined by the first recognition result. If the second recognition result is smaller than the second recognition result, a means for selecting the second recognition result as a final recognition result may be provided. According to such a configuration, the recognition probability of the user's voice and the recognition probability of the manager's voice are compared, and the recognition result with the higher recognition probability is adopted as the final recognition result. Thereby, a speech recognition system with higher recognition accuracy can be realized.

[0008] The supply means may further comprise means for displaying the first recognition result. This allows the administrator to repeat the user's voice while referring to the displayed recognition result, thereby reducing the workload of the administrator.

[0009] The speech recognition system comprises: a database;
Means for searching the database using the first recognition result generated by the first recognition means as a key, reading out corresponding data, and displaying the data. Thus, by recognizing one item of data (for example, a user ID or the like), data (for example, an address, a name, a telephone number, or the like) linked to the data can be automatically displayed. For this reason, it is possible to reduce the items that the user must utter, reduce the workload of the user, and realize an efficient system that can respond quickly.

[0010] The first recognition probability and the second recognition probability may be multiplied by a first weight value and a second weight value, respectively. Thereby, for example, it is possible to make adjustments so that the voice qualities of the user and the administrator become equal.

In a call center system according to a second aspect of the present invention, in a call center system in which an operator receives a telephone call from a user, a receiving means for receiving the voice of the user, and a receiving means for receiving the voice of the user. A first recognition unit that analyzes a voice of the user and generates a first recognition result and a first recognition probability for the first recognition result; and the first recognition unit that is generated by the first recognition unit. A means for presenting and displaying the recognition result to the operator, a means for supplying the voice of the user to the operator, an input means for inputting the voice of the operator, and a voice of the operator input by the input means Analyzing means for generating a second recognition result and a second recognition probability for the second recognition result; and Generated first and a means for the second recognition result is determined whether or not match, the first and second
If it is determined that the recognition results match, the first and second
Is determined as a final recognition result, and the first
Means for comparing the first and second recognition probabilities and selecting a final recognition result when it is determined that the first and second recognition results do not match.

According to such a configuration, recognition processing is performed on both the user's voice and the operator's voice, and when the recognition results are different, the higher recognition probability is used as the final recognition result. adopt. As a result, a call center system with higher recognition accuracy can be realized. Further, since the operator can repeat the user's business while referring to the displayed recognition result, the operator's work load can be reduced.

[0013] A voice recognition method according to a third aspect of the present invention includes a receiving step of receiving a voice of a user;
Analyzing the voice received by the receiving step,
A first recognition step of generating a recognition result of the first recognition result and a first recognition probability for the first recognition result; a repetition requesting step of supplying the voice to an administrator to request a repetition of the voice;
A second recognition step of analyzing a voice read back by the manager to generate a second recognition result and a second recognition probability for the second recognition result; and the first and second recognition means Determining whether or not the first and second recognition results generated by the first and second recognition processes match; and, if it is determined that the first and second recognition results match, the first and second recognition processes are performed. A result is determined as a final recognition result, and when it is determined that the first and second recognition results do not match, the first and second recognition results are determined.
And selecting a final recognition result by comparing the recognition probabilities of the two.

According to such a configuration, recognition processing is performed on both the user's voice and the manager's voice, and when the recognition results are different, the higher recognition probability is used for the final recognition. As a result, the accuracy of speech recognition can be improved.

A recording medium according to a fourth aspect of the present invention is a computer-readable recording medium that stores a program for causing a computer to function as a voice recognition system that accepts a business message from a user. Receiving means for receiving the user's voice, analyzing the user's voice received by the receiving means, and providing a first recognition result and a first recognition result for the first recognition result. First recognition means for generating a recognition probability;
Supply means for supplying the voice of the user to a manager, input means for inputting the voice of the manager, analyzing the voice of the manager input by the input means, and obtaining a second recognition result and the second A second recognition unit that generates a second recognition probability for the recognition result of the first determination unit, and determines whether the first and second recognition results generated by the first and second recognition units match. Determining means for determining the first and second recognition results as a final recognition result when determining that the first and second recognition results match; determining the first and second recognition results; If it is determined that does not match, a program for functioning as selection means for comparing the first and second recognition probabilities and selecting a final recognition result is recorded.

A recording medium according to a fifth aspect of the present invention is a computer-readable recording medium storing a program for causing a computer to function as a call center system that accepts telephone calls from a user by an operator. Receiving means for receiving the voice of the user, analyzing the voice of the user received by the receiving means, and providing a first recognition result and the first
First recognition means for generating a first recognition probability with respect to the recognition result, means for presenting the first recognition result generated by the first recognition means to the operator, and voice of the user Means for supplying to the operator, input means for inputting the voice of the operator, analyzing the voice of the operator input by the input means, a second recognition result and a second recognition probability for the second recognition result; Second recognition means for generating the first recognition result, means for determining whether or not the first and second recognition results generated by the first and second recognition means match, the first and second recognition means When it is determined that the results match, the first and second recognition results are determined as final recognition results, and when it is determined that the first and second recognition results do not match, the first and second recognition results are determined. Comparing the second recognition probability and final Recording the Do means for selecting a recognition result, a program to function as a.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A speech recognition system according to an embodiment of the present invention will be described below with reference to the drawings, taking as an example a speech recognition system installed at a support center for receiving inquiries regarding products, repair requests, and the like by telephone. I do.

As shown in FIG. 1, the speech recognition system includes a line control unit 11, an operator operation unit 12, a speech recognition unit 13, a main control unit 15, a database 17,
Is provided. The line control unit 11 receives voice from the user's telephone via a public line and transfers the voice to the voice recognition unit 13. The line control unit 11 connects / disconnects the line. The operator operation unit 12 includes a microphone, a keyboard, a speaker, a display device, and the like.

The voice recognition unit 13 includes an A / D conversion unit that converts analog voice data from the line control unit 11 and the operator operation unit 12 into digital data. The voice recognition unit 13 converts the voice data into digital data by an A / D conversion unit and performs a recognition process. The recognition method is arbitrary, such as a method using a probabilistic model, a method using a linear programming, a method based on an experiment, and the like, as long as a recognition result and a correct probability (likelihood) of the result can be obtained. Voice recognition unit 1
3 transmits the recognition result and the likelihood to the main control unit 15.

The main control unit 15 performs a process of comparing the recognition result of the user's voice from the voice recognition unit 13 with the recognition result of the operator's voice, and controls the entire voice recognition system. Further, when the user pronounces the user ID, the main control unit 15 refers to the database 17 to read out information of the corresponding user (user information), and displays the information on the display device of the operator operation unit 12. For example, as shown in FIG. 2, the database 17 stores user information such as a user name, an address, a telephone number, and the like using a user ID assigned to the user as a key.

The present system presents the recognition result of the user's voice to the operator as a provisional recognition result, reduces human error by the operator, and recognizes the recognition result of the user's voice and the recognition of the operator's voice. This is a speech recognition system that can improve the recognition accuracy by comparing (matching) results. Next, the flow of processing in this system will be described with reference to FIG.

First, the line controller 11 of the speech recognition system
Receives a telephone call (call) from the user via the public line (P1). In response to this, the main control unit 15 transmits a guidance for prompting input (utterance) of a message such as "What kind of request is it?" (P2). The user responds to this guidance and asks "Please repair the device."
And the like (P3). The main control unit 15 supplies the user's voice to the operator via the speaker of the operator operation unit 12 in response to the reception of the voice (P4).
At the same time, it instructs the voice recognition unit 13 to recognize the received voice.

The voice recognition unit 13 recognizes the voice of the user according to the instruction, and outputs the recognition result and the likelihood to the main control unit 15.
Send to The main control unit 15 displays a screen corresponding to the recognition result (for example, a failure repair reception screen when the recognition result is “failure repair”) on the display device of the operator operation unit 12 (P5). The operator repeats the item indicated by the voice from the user, such as "It is a repair." here,
If the operator determines from the display screen that the recognition result is correct, the operator can repeat the song while referring to the display screen (for example, a failure repair reception screen). The voice of the operator is transmitted to the main control unit 1 via the microphone of the operator operation unit 12.
5 (P6).

The main control unit 15 supplies the voice of the operator to the user via the line control unit 11 and the public line (P
At the same time, the voice recognition unit 13 is instructed to recognize voice. The voice recognition unit 13 recognizes the voice of the operator according to the instruction, and transmits the recognition result and the likelihood to the main control unit 15.

Here, the main control unit 15 compares the recognition result and the likelihood of the voice of the user with the recognition result and the likelihood of the voice of the operator, and generates a final recognition result. This comparison processing will be described with reference to FIG. Since the voice of the user is transmitted via the telephone line, it is expected that the recognition result of the voice of the operator will have higher recognition accuracy than the recognition result of the voice of the user. In consideration of this, in this comparison processing, each recognition result is weighted. The method of obtaining the value of the weight (weight) corresponding to each recognition result is arbitrary, and for example, may be a value experimentally obtained under various conditions, such as the type of telephone line and the voices of a plurality of operators. In this example, the weight on the recognition result of the user is W1, and the weight on the recognition result of the operator is W2.
And The likelihood of the user's recognition result is T1 and the likelihood of the operator's recognition result is T2.

First, the main controller 15 determines whether or not the recognition result of the user's voice matches the recognition result of the operator's voice (step S1). When it is determined that they match, the recognition result of the voice of the user recognized first is selected as the final recognition result (step S5), and the process ends. If it is determined that they do not match, the main control unit 15 determines the likelihood (T1, T2) of each recognition result and the weight (W1,
W2) is calculated as the recognition accuracy (R1, R2) and compared (step S2).
2, S3). In this comparison, when the recognition accuracy R1 of the voice of the user is higher (R1> R2), the recognition result of the voice of the user is selected as the final recognition result (step S5), and the process is terminated. If the recognition accuracy R2 of the voice of the operator is higher (R1 <R2), the recognition result of the voice of the operator is selected as the final recognition result (step S4).

In the above comparison processing, when the operator's recognition result is selected as the final recognition result, the main control unit 15 changes the screen displayed on the display device of the operator operation unit 12 to recognize the operator's voice. The screen is changed to a screen corresponding to the result, and the recognition result of the voice of the user recognized first is corrected (P8). Further, when the recognition result of the user's voice is selected as the final recognition result, the display screen of the display device is not changed.

In this way, by recognizing the voices of both the user and the operator, comparing the recognition results, and selecting the one with the higher recognition accuracy, a more accurate recognition result can be obtained. . Further, by presenting the recognition result of the user's voice recognized first as a provisional recognition result to the operator, the operator can repeat the content of the user's voice with reference to the recognition result. As a result, human errors such as listening mistakes can be prevented, and the workload of the operator can be reduced.

In a case where an input (utterance) from the user is required a plurality of times in one operation, the above-described processing (P2 to P8 in FIG. 2) is repeated. For example, when the request from the user is a request for “failure repair”, the main control unit 15
Displays a screen having a plurality of input items as shown in FIG. 5 on the display device of the operator operation unit 12 and transmits guidance to the user to ask for a product name (P2). In response to this guidance, the usage utters the product name (P
3). In response to the reception of the voice, the main control unit 15 supplies the user's voice to the operator (P4) and instructs the voice recognition unit 13 to recognize the received voice. Upon receiving an instruction from the main control unit 15, the voice recognition unit 13 executes the product name recognition process, and outputs the recognition result and the likelihood to the main control unit 15.
Pass to.

The main controller 15 displays the recognition result in the product name input field B1 on the screen shown in FIG. 5 (P5). The operator repeats the user's voice (in this case, the product name) from the speaker (P6). At this time, if the operator determines that the displayed recognition result is correct, the operator can repeat the song while referring to the product name displayed in the product name input field B1 on the screen. The main control unit 15 supplies the repeated product name to the user (P7) and instructs the voice recognition unit 13 to recognize the voice. The voice recognizing unit 13 performs a process of recognizing the name of the repetitive product, and passes the recognition result to the main control unit 15.
The main control unit 15 performs a comparison process on the two recognition results, and determines a final recognition result. Here, if the two recognition results are different and the final recognition result is the recognition result of the product name uttered by the operator, the recognition result in the product name input field B1 on the display screen is corrected (P8). When the final result is determined as the initial recognition result, guidance (eg, user ID) for asking the next input item is sent to the user (P2).

In the present system, the main control unit 15
The database 17 is searched based on the recognition result received from the voice recognition unit 13, and the corresponding data can be read out and displayed on the operator operation unit. The main control unit 15
Items (key items) serving as keys for referring to the database 17 are stored in advance, and when the key items are input and recognized, the database 17 is referred to based on the recognition result.

The database 17 in the present system is
As shown in FIG. 2, the user ID is used as a key. User ID
User according to the guidance to request
Is uttered, the uttered user ID is the speech recognition unit 1
3 and passed to the main control unit 15. The main control unit 15 searches the database 17 and reads out each item (user name, address, and telephone number) corresponding to the recognized user ID, and, together with the user ID as the recognition result, for example, as shown in FIG.
In the input fields B2, B3, B4, and B5 shown in FIG.

The operator repeats the user ID, the repetitive user ID is recognized by the voice recognition unit 13, and the main control unit 15 performs a comparison process of the recognition result.
The final recognition result is determined. As the final recognition result, the initial recognition result (the user I uttered by the user
D) The user I who was uttered by the operator instead of the recognition result of D)
When the recognition result of D is selected, the database 17 is searched again using the user ID as a key. in this way,
By providing a database using input items as keys, items corresponding to the key items can be read and presented to the operator. As a result, the trouble of inputting data can be saved, and the processing time can be reduced.

Next, a specific processing flow in the present system will be described with reference to FIG. 6 as an example of a case where a telephone for a product order is received from a user C (user ID “A021”) and the business is processed. I will explain. First, a telephone call (call) from the user C is received (received) by the line control unit 11 (F).
1). In response, the main control unit 15 sends guidance such as "Please do your business" to the user (F2). Upon receiving this guidance, the user utters, for example, "I want to order a product." The voice of the user is notified to the operator from the speaker of the operator operation unit 12 and passed to the voice recognition unit 13 to execute the recognition process. After the completion of the recognition process, the recognition result and its likelihood are passed to the main control unit 15. The main control unit 15 displays a “commodity order reception screen” (FIG. 7) corresponding to the recognized “commodity order” on the display device of the operator operation unit 12 (F3).

The operator listens to the user's voice and refers to the screen to repeat the user's message, for example, "It's a product order." The voice of the operator is notified to the user via a public line, and is also passed to the voice recognition unit 13 to execute a recognition process. After the completion of the recognition process, the recognition result and its likelihood are passed to the main control unit 15 (F4). The main control unit 15 performs a comparison process of the recognition result of the voice of the user and the recognition result of the voice of the operator, and determines a final recognition result (F5).

When the two recognition results match, the main control unit 15 sends a guidance such as "Please enter the product name of the order" to the user in order to process the next input item (F).
6). In response to this guidance, the user utters, for example, "a cordless telephone" BB phone "". The voice of the user is notified to the operator from the speaker of the operator operation unit 12 and passed to the voice recognition unit 13 to execute the recognition process. After the completion of the recognition process, the voice recognition unit 13
Passes the recognition result and its likelihood to the main control unit 15. At this time, for example, an erroneous recognition result of “cordless telephone“ PP phone ”” and a likelihood “0.4” are obtained, and the main control unit 1
5 The main control unit 15 displays the recognized “cordless telephone“ PP phone ”” in the order product name input field B11 (F7).

The operator listens to the voice of the user, determines that the screen display is incorrect, and sets the "cordless telephone" BB
The phone says, "I repeat." The voice of the operator is notified to the user via a public line, and is also passed to the voice recognition unit 13 to execute a recognition process (F8). In this case, it is assumed that the correct recognition result “cordless telephone“ BB phone ”” and the likelihood “0.8” have been obtained and passed to the main control unit 15 (F9).

The main control unit 15 compares the values obtained by multiplying the respective likelihoods and the weights of the recognition results of the user's voice and the operator's voice. For example, when the weights for the recognition result of the user's voice and the recognition result of the operator's voice are set to “0.8” and “0.7”, respectively, the recognition accuracy R1 for the user's voice is “0. 32
(= 0.4 × 0.8) ”and the recognition accuracy R2 for the voice of the operator is“ 0.56 (= 0.8 × 0.7) ”.
Becomes The main control unit 15 determines that the recognition accuracy of the voice of the operator is higher (R1 <R2), and determines the final recognition result as “cordless telephone“ BB phone ””. In this case, the main control unit 15 corrects the contents of the order product name input column B11 on the screen of FIG. 7 (F9).

Next, the main control unit 15 sends guidance such as "Please enter your user ID" to the user (F10). In response to this guidance, the user must enter his or her ID (in this case,
"A021"). The voice of the user is notified to the operator from the speaker of the operator operation unit 12 and passed to the voice recognition unit 13 to execute the recognition process. After the completion of the recognition process, the recognition result and its likelihood are stored in the main control unit 1
Passed to 5.

The main control unit 15 determines that the input data (user ID) from the user is a key item of the database 17 and receives “A02” received from the voice recognition unit 13.
The database 17 is searched by using "1" as a key, and the data of the corresponding name, address, and telephone number are read. Main control unit 1
5 displays the recognition result “A021” in the user ID input field B12 and displays the read data in the name input field B13, address input field B14, and telephone number input field B15 (F11).

The operator is the user I who spoke
Listening to D and referring to the screen, "A021."
And repeat the user ID of the user. The voice of the operator is notified to the user via a public line, and is also passed to the voice recognition unit 13 to execute a recognition process. After the completion of the recognition process, the recognition result and its likelihood are passed to the main control unit 15 (F12).

The main control unit 15 executes a comparison process between the recognition results of the user's voice and the operator's voice to determine the final recognition result (F13). If the two recognition results match, the main control unit 15 considers that all necessary input items have been entered on the product order reception screen, and transmits guidance such as “Thank you” to the user (F1).
4) Instruct the line controller 11 to disconnect the line.
Upon receiving this instruction, the line control unit 11 disconnects the line and ends the processing (F15).

In this way, by performing recognition processing on both the user's voice and the operator's voice and comparing the recognition results, a more accurate recognition result can be obtained. Also, by recognizing the voice uttered by the user, displaying the recognition result on the screen and presenting it to the operator, the operator can repeat the song while referring to the screen display. Can be.

Before the next item is input, a guidance requesting confirmation of the user, such as "Are you sure?", Is transmitted. If the user transmits a voice such as "Yes." Moving to the next input, for example, when a voice saying “No.” is transmitted, the user may be prompted to re-utter the item. At this time, the regenerated voice may be recognized by the voice recognition unit 13 or may be input by the operator from the keyboard of the operator operation unit 12.

The operator may give guidance from the microphone of the operator operation unit 12.

Note that the data stored in the database is not limited to the data described above, and various data may be stored. This speech recognition system may include a plurality of such databases.

In the above description, in the comparison processing,
Although the recognition accuracy obtained by multiplying the likelihood by the weight is compared, the comparison process may be performed using only the likelihood.

In the comparison processing shown in FIG. 4, a step of comparing the recognition degrees R1 and R2 and, when they match, selecting one of them may be further provided. This selection may be performed by the operator or the system may be performed according to a preset selection criterion.

The speech recognition system of the present invention can be realized by using a general computer system without using a dedicated system. For example, a medium (floppy disk, CD-ROM) storing a program for executing the above-described operation in a computer including an A / D converter for converting analog data into digital data and a microphone for inputting an operator's voice. Etc.) to install the program, a speech recognition system that executes the above-described processing can be configured. In the case where the above-described functions are realized by the OS sharing or the OS and the application being jointly used, only the part other than the OS may be stored in the medium.

The medium for supplying the program to the computer may be a communication medium (a medium that temporarily and fluidly stores the program, such as a communication line, a communication network, or a communication system). For example, the program may be posted on a bulletin board (BBS) of a communication network and distributed via the network. Then, by starting this program and executing it in the same manner as other application programs under the control of the OS, the above-described processing can be executed.

[0051]

As described above, according to the present invention,
Recognition processing is performed on both the user's voice and the manager's voice, and when the recognition results are different, the one with a higher recognition probability is adopted as the final recognition result. Thereby, more accurate voice recognition can be realized.

[Brief description of the drawings]

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

FIG. 2 is a diagram showing a structure of data stored in a database.

FIG. 3 is a diagram for explaining a processing flow in the present system.

FIG. 4 is a flowchart illustrating a comparison process.

FIG. 5 is a diagram showing a screen displayed on a display device of an operator operation unit when a failure repair is received.

FIG. 6 is a diagram for explaining processing when an order for a product is received from a user.

FIG. 7 is a diagram illustrating a screen displayed by a display device of an operator operation unit when an order for a product is received.

[Explanation of symbols]

 11 Line control unit 12 Operator operation unit 13 Voice recognition unit 15 Main control unit 17 Database

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁶ Identification code FI G10L 3/00 561 H04M 11/00 302 H04M 11/00 302 G06F 15/403 310Z

Claims (10)

[Claims] 1. A voice recognition system for receiving a business message from a user by voice, comprising: receiving means for receiving the voice of the user; analyzing the voice of the user received by the receiving means; A first recognition unit that generates a recognition result of the first recognition result and a first recognition probability for the first recognition result; a supply unit that supplies a voice of the user to a manager; and a voice of the manager. An input unit, a second recognition unit that analyzes a voice of the administrator input by the input unit, and generates a second recognition result and a second recognition probability for the second recognition result; Determining means for determining whether or not the first and second recognition results generated by the first and second recognition means match; and determining that the first and second recognition results match. Final recognition of the first and second recognition results Determining means for determining as a result, and selecting means for comparing the first and second recognition probabilities and selecting a final recognition result when it is determined that the first and second recognition results do not match. A speech recognition system comprising: 2. The method according to claim 1, wherein the selecting unit selects the first recognition result as a final recognition result when the first recognition probability is larger than the second recognition probability. The speech recognition system according to claim 1, further comprising: means for selecting the second recognition result as a final recognition result when the result is smaller than the second recognition result. 3. The speech recognition system according to claim 1, wherein said supply means further comprises means for displaying said first recognition result. 4. A speech recognition system comprising: a database;
4. A means for searching the database using the first recognition result generated by the first recognition means as a key, reading and displaying corresponding data, and displaying the data. The speech recognition system according to claim 1. 5. The voice quality of the user and the manager is improved by multiplying the first recognition probability and the second recognition probability by a first weight value and a second weight value, respectively. The speech recognition system according to any one of claims 1 to 4, further comprising an adjusting unit that adjusts the voice recognition to be equal. 6. The speech recognition system according to claim 1, wherein the first recognition probability and the second recognition probability have a first weight value and a second weight value, respectively.
Means for multiplying the first recognition probability multiplied by the first weight value when it is determined that the first and second recognition results do not match. Means for determining whether or not the second recognition probability multiplied by the second weight value coincides with the first recognition probability; and the second recognition probability multiplied by the first weight value and the second recognition probability. The second multiplied by the weight of
Means for selecting a recognition result having a higher recognition probability as a final recognition result when it is determined that the recognition probability does not match with the first recognition value; and a first recognition probability multiplied by the first weight value. Means for selecting one of the first and second recognition results as a final recognition result when it is determined that the second recognition probability multiplied by the second weight value matches the second recognition probability. The speech recognition system according to any one of claims 1 to 4, further comprising: 7. A call center system in which an operator accepts a telephone call from a user, a receiving means for receiving the voice of the user, and analyzing the voice of the user received by the receiving means to perform first recognition. First recognition means for generating a result and a first recognition probability for the first recognition result; means for presenting and showing the first recognition result generated by the first recognition means to the operator; Means for supplying the voice of the user to the operator; input means for inputting the voice of the operator; analyzing the voice of the operator input by the input means; A second recognition unit that generates a second recognition probability for the recognition result of the first and second recognition units; and determining whether the first and second recognition results generated by the first and second recognition units match each other. Means for determining, and when it is determined that the first and second recognition results match, the first and second recognition results are determined as a final recognition result, and the first and second recognition results are determined. Means for comparing the first and second recognition probabilities with each other and selecting a final recognition result when it is determined that the two do not match with each other. 8. A receiving step of receiving a user's voice, analyzing the voice received in the receiving step,
A first recognition step of generating a recognition result of the first recognition result and a first recognition probability for the first recognition result; a repetition requesting step of supplying the voice to a manager and requesting a repetition of the voice; A second recognition step of analyzing the repetition of the voice and generating a second recognition result and a second recognition probability for the second recognition result; and a second recognition step generated by the first and second recognition means. Determining whether or not the first and second recognition results match; and determining that the first and second recognition results match, finalizes the first and second recognition results. Determining the final recognition result by comparing the first and second recognition probabilities when it is determined that the first and second recognition results do not match; A speech recognition method, comprising: 9. A computer-readable recording medium storing a program for causing a computer to function as a voice recognition system that accepts a business message from a user, wherein the computer reads the voice of the user. Receiving means for receiving, first recognition means for analyzing a voice of the user received by the receiving means, and generating a first recognition result and a first recognition probability for the first recognition result; Supply means for supplying the voice of the user to the administrator; input means for inputting the voice of the administrator; analyzing the voice of the administrator input by the input means; and a second recognition result and the second A second recognition unit that generates a second recognition probability for the recognition result; a determination unit that determines whether the first and second recognition results generated by the first and second recognition units match. Another means, determining means for determining the first and second recognition results as a final recognition result when it is determined that the first and second recognition results match, the first and second recognition results If it is determined that the two do not match, a selection unit that compares the first and second recognition probabilities to select a final recognition result. A computer-readable recording medium recording a program for causing the program to function as: 10. A computer-readable recording medium which records a program for causing a computer to function as a call center system for receiving a telephone call from a user by an operator, the computer receiving the user's voice. Receiving means, analyzing the voice of the user received by the receiving means, generating a first recognition result and a first recognition probability for the first recognition result, a first recognition means, Means for presenting the first recognition result generated by the recognition means to the operator, means for supplying the user's voice to the operator, input means for inputting the voice of the operator, input by the input means Analyzing the voice of the operator, and generating a second recognition result and a second recognition probability for the second recognition result. A second recognition unit, a unit for determining whether the first and second recognition results generated by the first and second recognition units match, the first and second recognition results match. If it is determined, the first and second recognition results are determined as final recognition results. If it is determined that the first and second recognition results do not match, the first and second recognition results are determined. Means for comparing the recognition probabilities and selecting the final recognition result; and a computer-readable recording medium on which a program for functioning as a computer is recorded.