JP7163845B2 - Information processing device and program - Google Patents

Description

The present disclosure relates to an information processing device and a program.

Conventionally, there is a technique for converting speech into text using speech recognition technology. Specifically, in industrial fields such as call centers where voice communication is performed, such technology is used to improve operational efficiency.

For example, there is a technique of performing speech recognition simultaneously with a plurality of speech recognition engines and selecting a recognition result based on the reliability of each speech recognition engine (see Patent Document 1).

Also, there is a technique for selecting which one of the speech recognition results obtained by a plurality of speech recognition engines is to be adopted (see Patent Document 2).

JP 2017-76139 A JP 2010-85536 A

The recognition result of voice recognition is used to create correct text in voice call business. In the case of a call center, the correct text is, for example, a text summarizing the content of an inquiry from a user. In the past, correct texts were created manually while listening to voices, which required a lot of work time. In some cases, a correct text is created while referring to the results of speech recognition using a speech recognition engine for creation work. In this case, since the same company's engine is used, there is also the problem that the same error occurs many times, which is troublesome.

When using a speech recognition engine, there are methods of parallel processing multiple recognition methods and selecting recognition results that are close to the correct answer. There was a problem.

An object of the present disclosure is to provide an information processing apparatus and a program capable of easily and efficiently correcting and registering recognized important parts.

An information processing device according to the present disclosure is an information processing device for correcting and registering the result of speech recognition, and is an information processing device for correcting and registering the result of speech recognition, performing speech recognition and storing text. searching for the text that matches a predetermined rule based on each of a plurality of speech recognition units stored in the storage unit and the plurality of the texts stored in the storage unit, and storing the searched text as an important text an important part narrowing-down unit that stores the important text in a unit, reads out from the storage unit, and displays the important text; and reads from the storage unit the important text selected from the important texts displayed by the important part narrowing-down unit. and display the correct text, which is a text corrected based on the displayed important text, in a first form, which is a predetermined form, and store the correct text entered in the first form. and an information registration unit for recognizing the correct text and inputting information in a second form, which is a predetermined form, according to predetermined conditions.

Also, a program of the present disclosure is a program for causing a computer to function as each part of the information processing apparatus described in the present disclosure.

According to the information processing device and the program of the present disclosure, it is possible to easily and efficiently correct and register recognized important parts.

1 is a block diagram showing an example of the configuration of an information processing apparatus according to an embodiment; FIG. 1 is a schematic block diagram of a computer functioning as an information processing apparatus of this embodiment; FIG. It is a figure which shows an example of the screen interface output by the display part. FIG. 10 is a diagram showing an example when an area containing text other than important text is selected; 4 is a flowchart of a processing routine executed by an information processing device; FIG. 10 is a diagram showing an example of a case where a button for switching between a correction mode and a registration mode is provided in an important part narrowing down area;

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail. Note that the information processing apparatus according to the present embodiment is used in the operation of a call center, and the case where it is operated by an operator will be described as an example. In operation, there are a correction mode in which correction work is performed and a registration mode in which correction is not performed. In this embodiment, it is assumed that the correction mode is selected in advance.

<Configuration of Information Processing Apparatus According to Embodiment of the Present Invention>

FIG. 1 is a block diagram showing an example of the configuration of an information processing apparatus 10 of this embodiment. The information processing apparatus 10 having the configuration shown in FIG. 1 can be configured by a computer including a CPU, a RAM, and a ROM storing programs and various data for executing each processing routine to be described later.

For example, the information processing device 10 can be realized by a computer 50 shown in FIG. The computer 50 includes a CPU 51 , a memory 52 as a temporary storage area, and a non-volatile storage section 53 . The computer 50 also includes an input/output interface (I/F) 54 to which an input/output device (not shown) is connected, and a read/write (R/W) unit 55 for controlling reading and writing of data to and from a recording medium. Prepare. The computer 50 also has a network I/F 56 connected to a network such as the Internet. The CPU 51 , memory 52 , storage unit 53 , input/output I/F 54 , R/W unit 55 and network I/F 56 are connected to each other via a bus 57 .

The storage unit 53 can be implemented by a hard disk drive (HDD), solid state drive (SSD), flash memory, or the like. A program for causing the computer 50 to function is stored in the storage unit 53 as a storage medium. The CPU 51 reads out the program from the storage unit 53, develops it in the memory 52, and sequentially executes the processes of the program. Information of the recognition result storage unit 22 is stored in the storage unit 53 .

The above is the description of an example of the electrical configuration of the computer in FIG.

Functionally, as shown in FIG. 1, the information processing apparatus 10 includes each of a plurality of speech recognition units 20, a recognition result storage unit 22, an important part narrowing down unit 24, a correction operation unit 26, An information registration section 28 and a display section 30 are provided. Note that the recognition result storage unit 22 is an example of the storage unit of the present invention.

Here, an example of a screen interface output by the display unit 30 will be described. FIG. 3 is a diagram showing an example of a screen interface output by the display unit 30. As shown in FIG. As shown in FIG. 3, in the screen interface, an important part narrowing-down area 24A, a correction work area 26A, and an information registration area 28A are integrally formed on the display screen. Each area corresponds to the important part narrowing down section 24, the correction work section 26, and the information registration section 28, respectively. The information processing apparatus 10 reflects various texts of processing results of each processing unit in each area of the screen interface. With such a screen interface, by dividing the screen into three areas for each task and displaying it, various tasks can be completed on one screen. Therefore, for example, an operator at a call center or the like can efficiently record a history of conversations in a short period of time. A specific method will be described later in the process of each unit corresponding to each area.

Each of the speech recognition units 20 is a speech recognition engine that performs dialogue speech recognition. The speech recognition engine performs speech recognition using different techniques for each speech recognition unit 20 . Each of the speech recognition units 20 performs speech recognition and stores the text in the recognition result storage unit 22 . Speech recognition is grouped by the speaker's utterances, if identifiable by the speech recognition engine, and the text is labeled with the speaker's identification.

The recognition result storage unit 22 stores the labeled text recognized by each of the speech recognition units 20 . In addition, the important text searched by the important part narrowing-down unit 24 is stored. Also, the important text corrected by the correct text corrected by the correction working unit 26 is stored. Also, the recognition rate of each speech recognition engine processed by the correction work unit 26 is stored.

The important part narrowing-down unit 24 searches for texts that match a predetermined rule based on the multiple texts stored in the recognition result storage unit 22 . The important part narrowing-down unit 24 stores the retrieved text as important text in the recognition result storage unit 22, reads the text from the recognition result storage unit 22, and displays the important text in the important part narrowing-down area 24A. The important part narrowing-down unit 24 also accepts selection of important texts. The selection may be made by clicking the important text in the balloon 24B or by selecting the text by drag-and-drop or the like.

The important part narrowing-down unit 24 extracts utterances and parts within utterances that are judged to be important parts from the dialogue, and as shown in FIG. Display important text for each speaker identified. In FIG. 3, the icon on the right side represents the user who made the inquiry, the icon on the left side represents the operator's statement, and an important text is displayed in the balloon 24B. As for the predetermined rule for judging an important part, it is sufficient to learn a rule or a discriminant model for judging an important part by adopting or combining various methods. Predetermined rules include, for example, extracting sentences containing specific keywords, extracting relevant sentences after understanding the context using natural language analysis technology, extracting relevant sentences by omitting unnecessary sentences, etc. is mentioned.

Also, there may be cases where there is no important part in the displayed important text. The important part narrowing-down area 24A is provided with an area 24C containing texts other than the important texts displayed with "...". The area 24C includes the text recorded between the upper important text and the lower important text on the dialogue chronological order.

FIG. 4 is a diagram showing an example when the area 24C including text other than the important text is selected. By selecting the area, it is possible to display and extract other utterances that have not been picked up as important text. Sentences that are determined to be important and extracted can be added to the important part narrowing down area 24A as important texts in balloons 24B. It should be noted that a speech balloon of another statement may be displayed as a pop-up or the like.

The correction work section 26 reads important texts selected from the important texts displayed in the important part narrowing-down area 24A from the recognition result storage section 22, and displays information about the important texts selected in the display area 26B of the correction work area 26A. display. The correction work section 26 receives input of correct text from the correction work form 26C in the correction work area 26A. The correct text is information about the important text displayed in the display area 26B, which is text corrected based on the speech recognition results of the speech recognition engine displayed in parallel. The correction work unit 26 accepts pressing of the correct text registration button, and stores the correct text as a corrected important text in the recognition result storage unit 22 . The operator performs the input and registration work on the correction work form 26C. Further, the correction work unit 26 reads out the corrected important text with the correct text from the recognition result storage unit 22 and reflects it in the display as the corrected important text in the important part narrowing down area 24A. Note that the correction work form is an example of the first form.

The correction work area 26A is an area in which it is possible to correct an important text judged to be an important part by the important part narrowing-down unit 24, that is, if there is an erroneous recognition of a sentence of the selected important text. be. For the selected important text, it is possible to display the results of multiple speech recognition engines side by side in the display area 26B. It is possible to automatically generate a correct sentence in the correction work form 26C by regarding the portion where the results of a plurality of speech recognition engines match as the correct answer. In this manner, the correction work area 26A is an area in which the operator can manually check the correct sentence draft, correct it as necessary, and complete the correct sentence.

In addition, when the operator completes the correct text in the correction work form 26C and presses the correct text registration button, the correction work unit 26 performs a , the recognition rate for each speech recognition engine is calculated and stored in the recognition result storage unit 22 . The recognition result can also be displayed in the display area 26B. The recognition rate may be calculated from the match rate or the like when the original text of the speech recognition engine is compared with the input correct text. The recognition rate can be used to evaluate speech recognition engines.

Further, when the result of correction of the correct text is reflected in the important part narrowing-down area 24A, the fact that the important text has been corrected may be indicated by an annotation, a change in font color, or the like. By reflecting correction targets in an identifiable format in this manner, the operator can grasp which important text has been corrected when there are multiple corrections.

The information registration unit 28 recognizes the corrected important text in the important part narrowing down area 24A, and allows the information to be entered in the information registration forms 28B to 28E in the information registration area 28A according to predetermined conditions. The information registration unit 28 accepts pressing of the registration button and stores the registration information in the recognition result storage unit 22 .

The information registration forms 28B to 28E are each provided with a category. For example, the information registration forms 28B and 28C can be defined as "free entry", 28D as "item", and 28E as "answer". In the information registration area 28A, using the important texts displayed in the important part narrowing down area 24A, sentences automatically generated for each category and sentences copied from the important texts are entered into information registration forms 28B to 28E prepared in advance. By reflecting the information and applying it to the content of the user's inquiry, it is possible to easily perform information registration. The automatic generation of sentences may be performed automatically when the important text is selected, or may be performed by clicking on the information registration form. Copying may also be performed by double-clicking on important text.

In the automatic generation of sentences, for example, when the speech balloon 24B in the important part narrowing down area 24A is a long sentence of 10 lines or more, unnecessary words are removed to simplify the sentence. You may use an automated writing tool designed to be done in . This is because the important text displayed in the balloon 24B is spoken language and may include many unnecessary words.

<Operation of the information processing apparatus according to the embodiment of the present invention>

Next, operation of the information processing apparatus 10 will be described with reference to FIG. FIG. 5 is a flowchart of a processing routine executed by the information processing apparatus 10. As shown in FIG. The processing routine is executed, for example, at the same time as the user starts interacting with the call center.

In step S<b>100 , each speech recognition unit 20 performs speech recognition and stores labeled text in the recognition result storage unit 22 .

In step S<b>102 , the important part narrowing-down unit 24 searches for text that matches a predetermined rule based on the multiple texts stored in the recognition result storage unit 22 .

In step S104, the important part narrowing-down unit 24 stores the retrieved text as important text in the recognition result storage unit 22, reads it from the recognition result storage unit 22, and displays the important text in the important part narrowing-down area 24A.

In step S106, the important part narrowing-down unit 24 receives selection of important texts from the important texts displayed in the important part narrowing-down area 24A.

In step S108, the correction work unit 26 reads the important text selected from the important texts displayed in the important part narrowing down area 24A from the recognition result storage unit 22, and selects it in the display area 26B of the correction work area 26A. Display information about important text.

In step S110, the correction work section 26 receives input of correct text from the correction work form 26C in the correction work area 26A. Based on the displayed contents, the operator enters the correct text into the correction work form 26C.

In step S112, the correction working unit 26 accepts pressing of the correct text registration button, and stores the correct text in the recognition result storage unit 22 as corrected important text.

In step S114, the correction working unit 26 reads the important text corrected with the correct text from the recognition result storage unit 22, and reflects the corrected important text in the important part narrowing down area 24A. do.

In step S116, the correction unit 26 calculates the recognition rate for each of the speech recognition engines displayed in parallel in the display area 26B, and stores it in the recognition result storage unit 22. FIG.

In step S118, the information registration unit 28 recognizes the corrected important text in the important part narrowing down area 24A, and according to predetermined conditions, inputs information to the information registration forms 28B to 28E in the information registration area 28A. accept.

In step S120, the information registration unit 28 accepts pressing of the registration button, stores the registration information in the recognition result storage unit 22, and ends the process.

As described above, the information processing apparatus 10 according to the present embodiment can easily and efficiently correct and register recognized important parts.

The present invention is not limited to the above-described embodiments, and various modifications and applications are possible without departing from the gist of the present invention.

In the above-described embodiment, the case where the correction mode is selected in advance has been described as an example, but the present invention is not limited to this. For example, in the important part narrowing down area 24A, a screen interface for switching between selecting a correction target and registering information may be used. FIG. 6 is a diagram showing an example in which a button for switching between the correction mode and the registration mode is provided in the important part narrowing-down area 24A. As shown in FIG. 6, the mode is selected by the mode selection area 24D. When the correction mode is selected, when important text is selected in the important part narrowing down area 24A, the important text is reflected in the display area 26B of the correction work area 26A and the correction work form 26C. When the registration mode is selected, when an important text is selected in the important part narrowing down area 24A, it is automatically reflected in the information registration forms 28B to 28E according to predetermined conditions. Further, the operator can register information by copying the important text in the important part narrowing down area 24A to the information registration forms 28B to 28E. Further, when the registration mode is selected, the correction work area 26A may be hidden. By allowing the mode to be switched in this way, the operator can easily perform the registration work on one screen.

Further, in the specification of the present application, an embodiment in which the program is pre-installed has been described, but it is also possible to store the program in a computer-readable recording medium and provide it.

10 Information processing device 20 Voice recognition unit 22 Recognition result storage unit 24 Important part narrowing unit 26 Correction operation unit 28 Information registration unit 30 Display unit

Claims (7)

An information processing device for correcting and registering voice recognition results,
each of a plurality of speech recognition units that perform speech recognition and store text in a storage unit;
Based on the plurality of texts stored in the storage unit, the texts matching a predetermined rule are searched, the searched texts are stored in the storage unit as important texts, and the important texts are read out from the storage unit. an important part narrowing section for displaying text;
The important text selected from the important texts displayed by the important part narrowing-down unit is read from the storage unit and displayed, and a correct text, which is a text corrected based on information related to the displayed important text, is displayed. a correction work unit that prompts the user to enter a first form, which is a predetermined form, and stores the correct text entered in the first form in the storage unit;
an information registration unit that recognizes the correct text and inputs information in a second form, which is a predetermined form, according to predetermined conditions;
Information processing equipment including. further comprising a display;
3. The display section is formed so as to integrally display a display area corresponding to the important part narrowing-down section, a display area corresponding to the correction work section, and a display area corresponding to the information registration section. 1. The information processing device according to 1. 3. The information processing apparatus according to claim 1, wherein the correction operation unit reads the selected important text and a plurality of texts corresponding to the selected important text from the storage unit and displays them. 4. The information processing according to any one of claims 1 to 3, wherein the recognition rate of each of the speech recognition units is measured based on the correct text registered by the correction work unit and registered in the storage unit. Device. The important part narrowing-down unit
Based on selection of an area containing text other than the important text, candidate texts other than the important text are displayed, and text selected from the candidate texts is stored in the storage unit as the important text. Item 5. The information processing apparatus according to any one of Item 4. the predetermined conditions include conditions related to automatic reflection of the text;
A plurality of the second forms are provided for each category,
6. The information registration unit reflects the information candidate in the second form of the category corresponding to the important text and the correct text according to the predetermined condition. The information processing device according to . A program for causing a computer to function as each unit of the information processing apparatus according to any one of claims 1 to 6.

Images