KR20090094577A - Method for evaluating spoken ability through computer-lead speech recognition - Google Patents

Abstract

A method for evaluating computer-lead type speaking ability by using voice recognition is provided to calculate a score by a computer, thereby objectively, reliably and immediately evaluating speaking ability about a specific language of an examinee. A background image, a character and a background sound are outputted by a computer(S1). An interaction voice signal for speaking evaluation is outputted from a speaker(S2). An examinee gives an answer about the interaction voice signal for evaluation through a microphone(S3). It is determined whether the voice signal that the examinee answers is recognized(S4). If the voice signal is recognized, an answer of the examinee is stored in a database.

Description

METHOD FOR EVALUATING SPOKEN ABILITY THROUGH COMPUTER-LEAD SPEECH RECOGNITION}

The present invention relates to a method of evaluating computer-driven speaking ability using speech recognition, and in particular, when the examinee presents a conversation, discussion, or opinion in a direction in which the computer presents the speaking ability of a specific language, The present invention relates to a method of evaluating computer-driven speaking ability using speech recognition, which can evaluate the examinee's speaking ability in a specific language on the fly by calculating scores according to stored evaluation criteria.

Conventional speaking evaluation, which has been widely used in the past, has been mainly conducted by humans and has been evaluated by humans. However, this evaluation has a problem that it takes a lot of time, and also objective evaluations, scorers' reliability, and a large number of examinees can be evaluated on the fly. There was no problem.

In order to improve such a problem, a method of recording an examinee's speaking attempt through a computer and delivering the score to a grader has been widely utilized.

However, this evaluation also has a problem that the score of the examinee is still determined by the scorer's supervision and the scorer selection and the scoring period takes a long time, and the examinee must only answer the question for the speech evaluation manually. There was a problem of closedness because the examinee expressed his opinion and lacked negotiation with the other party for dialogue. In addition, the grading criteria were also composed of a simple grading system, and there was a problem in that it was not possible to solve the unreasonable point of determining the examinee's score by selecting the examinee's performance by the scorer's feeling.

In order to solve this problem, a recent computer-aided speech assessment is performed by the examinee by simply reading or following the information presented by the computer, repeating the sentence, and storing the speaking attempts with short answers. In addition, indirect speaking evaluations are shown in which scores are determined depending on whether or not a sentence is in error, but such speaking evaluation is divided into grades based on the examinee's pronunciation or sentence error. In addition, there was a problem of evaluating speaking ability only by pronunciation accuracy. In the evaluation of speaking ability through the computer, the disadvantage that the evaluation cannot be made considering the speaking ability such as the logic of the examinee's conversation, the fluency, the use of the speaking strategy and the use of the proper vocabulary has been raised.

In order to solve these problems, the applicants of the present invention have been registered in 2000 and 4 cases (0287227,0287228,0296271,0296272), and a patent application on June 26, 2007 (Patent Application No. 10-2007-0063336) In a computer-driven conversational device and method, a computer is used to induce 1: 1 or a large number of language conversations by the computer, and further develops that the computer is designed to induce a conversation and the examinee is designed to communicate according to the computer's induction. A system for evaluating even the capability was established through the present invention.

Conventional speaking evaluation is based on the overall grading criteria for the examinee's feelings of speaking attempts and partial grading criteria, such as pronunciation, fluency, grammatical errors, and vocabulary selection. It was true that objective scoring was difficult, for example, it took a considerable amount of time, and the scores vary depending on the physical condition and mood of the grader. In particular, in the case of human grading, the scores are highly dependent on the overall grading criteria that give the overall impression to the examinees.In order to score based on the partial grading criteria, the scores of the examinees are recorded and the scores are calculated according to the detailed grading criteria. It took a lot of time to follow the crowd was true. When humans scored, there was a problem that it was almost impossible to grade the scores by listening to the examinees and comparing them with the partial grading criteria.

Therefore, in the case of human grading, the grading depends on feelings, impressions, moods, and feelings of the examinee, so that there is a problem in that objectivity cannot be maintained. In addition, recent grading by computers tends to rely only on partial grading. There was a problem that can not maintain objectivity.

In addition, the scoring by the computer mechanically classifies the correct answer, wrong answer, grammatical error, etc. of the examinee's pronunciation, recognizes correct pronunciation and grammatically perfect sentences as correct answers, and recognizes sentences containing some errors as incorrect answers. Even in the case of the examinee who can communicate smoothly in a general conversation situation, the communication ability, the logic of the conversation, the use of proper vocabulary, the fluency, etc. are not considered. As a result, there was a problem of being judged with a low score.

SUMMARY OF THE INVENTION The present invention has been made to solve the above various problems, and compares the speaking input of the examinee with the evaluation criteria recorded in advance in the database of the data section to instantly evaluate the examinee's speaking ability of a specific language. The purpose of this study is to provide a computer-driven speech language learning ability evaluation method using speech recognition that can be derived.

It is another object of the present invention to consider the feelings, naturalness, and impression of the examinee's speaking attempt as an overall grading criterion in which the examinee's speaking attempt is stored in a database. Appropriateness, relevance of content, clarity of expression, fluency, etc.) are computer-driven speech using speech recognition that can be input into the database and summed up partial scores to quantify the total score. To provide a method for assessing abilities.

The present invention has been made to achieve the above object, and the computer-driven speech ability evaluation method using the speech recognition of the present invention includes a background image for outputting a background image, text and background sound for speech evaluation by a computer. Text / background sound output process; An evaluating dialogue voice signal output process for outputting a dialogue speech signal for speech evaluation from the speaker after the background image / text / background sound output process; An answering step in which the examinee answers the answer through the microphone after the voice signal is recognized in the evaluation dialogue voice signal output process; A voice recognition determining step of determining whether the examinee has recognized a voice signal answered by the examinee; An answer data search process of searching for an evaluation criterion answer data stored in an evaluation criterion database of the data part under the control of the controller when the answer of the examinee recognizes the voice signal answered by the examinee in the voice recognition discrimination; An answer similarity determination process for determining whether the evaluation reference answer data searched in the answer data search process and the examinee's answer are similar; A suitable answer evaluation step of evaluating an appropriate answer when the examinee's answer is similar to the evaluation criterion answer data retrieved from the evaluation criterion database of the data unit; And an evaluation result storage step of storing the evaluation result of the fit response of the fit response evaluation process in the evaluation result database of the data unit.

According to the computer-driven speech language evaluation method using the voice recognition according to the present invention, by comparing the examinee's speaking input with the evaluation criteria recorded in the database of the data section in advance, the speaking ability of the examinee's specific language is highly objective and reliable. In addition to being able to evaluate on the fly, the examinee's attempt to speak is an overall scoring criterion stored in the database of the data section, taking into account the feelings, naturalities, and impressions of the examinee's attempts to speak, and analyzing analytic scoring criteria Accuracy, adequacy of vocabulary selection, relevance of content, clarity of expression, fluency, etc.) can be entered into the database based on the native speaker's speech utterance data, and the partial scores can be summed to quantify the total score. Excellent effect.

In the description of the present invention, the configuration of the database for evaluating the speaking utterance ability of the examinee establishes a database based on a scoring standard which is an integrated type of comprehensive scoring and analytical scoring criteria. Each area of the grading criteria is assigned one point between each grade, from 0 to 5 points, and consists of a total of 0 to 60 points (which can be graded between grades depending on the nature and purpose of the test). Evaluation criteria of each area are as follows.

As a general grading standard, it considers the feelings, naturalness, impression, fluency, and logic of the examinee's attempt to speak in general, and analyzes the grading criteria (the accuracy or adequacy of the pronunciation, the grammatical accuracy, the relevance of the vocabulary selection, the relevance of the contents, the clarity of expression, Abundance of language using strategies such as fluency, task realization, vocabulary, grammar and transliteration to convey meaning

Comprehensive scoring

1) Natural speaking utterance of examinee: Evaluate conversational techniques, manners, etc.

2) the examinee's speaking logic: assessing the ability to consistently solve the subject of conversation

3) Fluency in conversation: evaluation in consideration of speaking strategies (speaking, respeaking, hesitating time) in conversation

Analytical scoring criteria

1) Task realization and time: Comparing the ability to respond to and solve a given conversation task, task realization time, etc. with native speaker average task realization time

2) Conversational Strategies (adequate conversational skills, verbalization strategies such as conjunctival to maintain a smooth conversation, rest period, etc.)

3) Grammar accuracy: Evaluation based on achievement of meaning transfer, not two-step evaluation of correct and incorrect answers

5) Pronunciation: Differentiated grades depending on the clarity of the examinee's words and whether they affect the communication from general native speaker's level compared to the native speaker's pronunciation.

6) Fluency: Comparison with native speakers' data based on the examinee's speech rate, rhythm and hesitation

7) Use of core vocabulary: Judging whether or not to convey meaning quickly by using simple and clear core vocabulary

Hereinafter, a computer-driven speaking language learning ability evaluation apparatus using speech recognition according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram schematically illustrating an apparatus for evaluating computer-driven speaking language learning ability using speech recognition according to an embodiment of the present invention, and FIG. 2 is a block diagram showing a detailed structure of a processor and a data unit in FIG. 1. to be.

1 and 2, a computer-driven speaking ability evaluation apparatus using voice recognition according to an embodiment of the present invention is a program storage unit for storing a program provided to examinees who want to receive an interactive language evaluation led by a computer. 10; A control unit 20 for controlling the answer in response to the searched answer for controlling the overall operation to evaluate the speaking ability according to the examinee's request; A processing unit 30 for recognizing a voice signal and converting a response voice signal, a background sound signal, and a background image signal from corresponding data and outputting the corresponding data; A data unit 40 for storing the recognized voice signal, the voice signal to be output, the background sound signal, and the background image signal in a database; A microphone 50 for inputting the examinee's voice response signal or voice question signal; A display unit 60 for outputting a background image and text for evaluation in various colors; An input unit 70 for inputting an evaluation request command and data information for an evaluation request for speaking; A speaker 80 for outputting a speech signal and a background sound signal for evaluating speech; The communication unit 90 communicates with other computers by wire or wirelessly.

In addition, as shown in FIG. 2, the processor 30 includes a voice recognition unit 31 for recognizing a voice signal and converting the voice signal into data; A voice output unit 32 which converts the retrieved and provided voice data into a voice signal and outputs the voice signal; A background sound unit 33 for converting the searched and provided background sound data into an audio signal; The background image unit 34 converts the searched and provided background image data or text data into an image signal.

In addition, the data unit 40 includes a voice recognition database 41 for storing voice signal data as data; A voice output database 42 for storing voice output data and outputting the retrieved voice data to the controller 20 for evaluation of the speaking ability; A background sound database (43) for storing background sound data for evaluating speaking ability, retrieving the stored background sound data, and outputting the stored background sound data to the controller (20); A background image database 44 for storing background image data for language ability, retrieving the stored background image data, and outputting the stored background image data to the controller 20; While storing the evaluation criteria (evaluation criteria) for evaluating the examinee's speech utterance ability, receiving the voice output data stored in the voice output database 42 and comparing the answer signal to output the comparison result to the control unit 20 An evaluation criteria database 45; It consists of a dialogue subject / topic database 46 which stores the dialogue subject data or the subject data, and the evaluation result database 47 which stores the examinee's evaluation result data.

Next, the operation and effects of the computer-driven speech language learning ability evaluation apparatus using the speech recognition according to the embodiment of the present invention configured as described above will be described.

First, in a computer-driven speaking language learning ability evaluation apparatus using the voice recognition of the present invention according to an embodiment of the present invention, an evaluation request signal for the examinee requesting to start language evaluation at the input unit 70 is transmitted to the control unit 20. When input, the control unit 20 loads a corresponding application for language evaluation from the program storage unit 10, and becomes a standby state for providing a language evaluation service to the examinee.

At this time, the control unit 20 monitors the communication unit 90 to determine whether another examination (or learning) computer or a management computer is connected to the communication unit 90, and as a result of the determination, the other examination computer or management computer When connected to a computer, the current operation status information and the operation information necessary for management are exchanged with each other, and the necessary information is periodically controlled as the speech evaluation progresses.

The controller 20 provides the speaking ability evaluation service from the voice output database 42, the background sound database 43, and the background image database 44 of the data unit 40 to provide the examinee ability evaluation service to the examinee. An audio signal including a background image signal, a background sound signal, and text data is searched for and provided to the processor 30.

The processing unit 30 transmits the background image signal, the background sound signal, and the text data provided from the control unit 20 to the background image unit 34, the background sound unit 33, and the audio output unit 32 of the processing unit 30, respectively. The background image unit 34 of the processing unit 30 converts the background image data into a background image signal (including a text signal), and the background sound data of the background unit 33 of the processing unit 30 performs the background. The audio signal is converted into an audio signal, and the text data, which is the audio data of the processor 30, is converted into an audio signal by TTS (TEXT TO SPEECH) technology.

The video signal and the text signal respectively converted by the processing unit 30 are output from the control unit 20 and output to the display unit 60 to display an image screen on the display unit 60. The background sound signal converted by the background sound unit 33 of FIG. 30 and the background sound signal converted by the sound output unit 32 of the processor 30 are provided to the speaker 80 to output sound and voice.

The control unit 20 monitors whether the examinee's voice signal is input from the microphone 50 and checks whether the examinee's voice signal is provided to the voice recognition unit 31 of the processing unit 30 and recognized.

In the above description, a predetermined ratio set for measuring similarity is a state in which the response voice signal input and recognized through the microphone 50 includes the noise of the surrounding environment, a change in the examinee's intonation, the use of similar words and phrases, and the examinee. The appropriate value is selected and evaluated in a repeated experiment in consideration of differences in voice signals due to age and gender.

3 is a flowchart schematically illustrating a method for evaluating computer-driven speaking ability using speech recognition according to an embodiment of the present invention. In the figure, S denotes a step.

As shown in FIG. 4, the method for evaluating computer-driven speaking ability using voice recognition according to an embodiment of the present invention outputs the background image, text and background sound for speech evaluation by the computer in step S1 to the display unit 60. (Background image / text / background sound output process), and proceeds to step S2 and outputs the dialogue voice signal for speech evaluation from the speaker 80 after the background image / text / background sound output process (dialogue audio signal output for evaluation). Step) After the voice signal is recognized in the evaluation dialogue voice signal output process, the process proceeds to step S3 where the examinee answers the response dialogue voice signal through the microphone 50 (answer process).

Next, the process proceeds to step S4 to determine whether or not the voice signal answered by the examinee has been recognized in the answering process (voice recognition determination process), and when the voice signal answered by the examinee is recognized by the voice recognition discrimination determination (YES). In step S5, the examinee's answer is retrieved from the evaluation criteria database 45 stored in the evaluation criteria database 45 of the data unit 40 under the control of the control unit 20 (response data search process), and the process proceeds to step S6. It is determined whether the evaluation criteria answer data retrieved in the answer data search process and the examinee's answer is similar (determining whether or not there is a response).

Next, in the process of determining whether or not the answer is similar, if the examinee's answer is similar to the evaluation criterion answer data retrieved from the evaluation reference database 45 of the data part 40 (YES), the process proceeds to step S7 and an appropriate answer is obtained. The evaluation is made (conformity answer evaluation process), the process proceeds to step S8, and the result of the conformance answer evaluation of the conformity answer evaluation process is stored in the evaluation result database 49 of the data section 40 (evaluation result storage process).

On the other hand, if the answer of the examinee in the process of determining whether or not the answer is similar to the evaluation criteria answer data retrieved from the evaluation criteria database 45 of the data section 40 (if NO), go to step S9 to the wrong answer Evaluate (wrong answer evaluation process). Proceeding to the evaluation result storage step of step S8, the incorrect answer evaluation result of the incorrect answer evaluation process is stored in the evaluation result database 49 of the data unit 40.

If the voice signal determined by the examinee is not recognized (NO) in step S4, the process returns to the background image, text / background sound output process of step S1, and returns the background image and the like. Text / Background sound output process The following process is repeated.

In the above description, the specific embodiments have been shown and described, but the present invention is not limited thereto, for example, by those skilled in the art without departing from the concept of the present invention. Of course, the design can be changed in various ways.

1 is a block diagram schematically illustrating an apparatus for evaluating computer-driven speaking language learning ability using speech recognition according to an embodiment of the present invention.

FIG. 2 is a block diagram illustrating a detailed structure of the processing unit and the data unit in FIG. 1.

3 is a flowchart schematically illustrating a method for evaluating computer-driven speaking ability using speech recognition according to an embodiment of the present invention.

          <Explanation of symbols for the main parts of the drawings>

10: program unit 20: control unit

30: processing unit 31: voice recognition unit

32: sound output unit 33: background sound unit

34: background image part 40: data part

41: voice recognition database 42: voice output database

43: background sound database 44: background image database

45: evaluation criteria database 46: conversation material, the topic database

47: evaluation results database 50: microphone

60: display unit 70: input unit

80: speaker 90: communication unit

Claims (3)

A background image, text, and background sound output process of outputting the speech image, the background image, the text, and the background sound to the display unit 60 by a computer; A dialogue voice signal outputting process for outputting a dialogue speech signal for speech evaluation in the speaker 80 after the background image / text / background sound output process; An answering process in which the examinee answers the answer through the microphone 50 after the speech signal is recognized in the evaluation dialogue voice signal output process; A voice recognition determining step of determining whether the examinee has recognized a voice signal answered by the examinee; In the case of recognizing the voice signal answered by the examinee in the voice recognition discrimination discrimination, the examinee's response is searched for the evaluation criteria response data stored in the evaluation criteria database 45 of the data unit 40 under the control of the controller 20. Retrieval process; An answer similarity determination process for determining whether the evaluation reference answer data searched in the answer data search process and the examinee's answer are similar; An appropriate answer evaluation step of evaluating a suitable answer when the examinee's answer is similar to the evaluation criterion answer data retrieved from the evaluation criterion database 45 of the data part 40; Evaluating the computer-aided speech ability using speech recognition, characterized in that it comprises a; storing the evaluation results of the appropriate response evaluation process of the appropriate response evaluation process in the evaluation result database 49 of the data unit 40; Way. According to claim 1, If the answer of the examinee in the process of determining whether or not the answer is similar to the evaluation criteria answer data retrieved from the evaluation criteria database 45 of the data portion 40 Goes to the wrong answer evaluation process and the wrong answer And after the wrong answer evaluation process, proceeds to the evaluation result storing process and stores the wrong answer evaluation result of the wrong answer evaluation process in the evaluation result database 49 of the data unit 40. A computer-driven speech assessment method using recognition.   The process of claim 1, wherein if the examinee does not recognize the voice signal answered by the examinee, the process returns to the background image / text / background sound output process and outputs the background image / text / background sound output. Computer-driven speaking ability evaluation method using the speech recognition, characterized in that the following process.

Images