Method for interactively scoring open short conversation by using computer

Abstract

The invention provides a method for interactively scoring an open short conversation by using a computer, which comprises the following steps of: collecting a voice database and training an acoustic model; acquiring expert knowledge and training a scoring model; extracting the characteristics of a sample voice signal and retrieving key contents; clustering voice contents when legal expert knowledge is not detected and performing content representation on clusters; and performing artificial check and supplementing and synthesizing the expert knowledge to obtain the content scores and feedback suggestions of all voice samples. Aiming at the characteristic that the expert knowledge difficultly completely covers the contents of a sample library in oral open questions, the expert scoring level is basically achieved in a man-machine interaction mode under the condition of consuming an extremely small amount of labor; meanwhile, suggestions about vocabulary usage and sentence pattern usage also can be provided for tested personnel in the evaluating process to guide the tested personnel to correct, and the method has high application value in a large-scale examination system and is not limited by language.

Description

technical field [0001] The invention belongs to the fields of speech recognition, language evaluation, and computer-aided language learning. It proposes a method for interactively scoring open short dialogues by using a computer, and specifically relates to speech keyword retrieval technology, speech content clustering technology, and high-precision phonemes. recognition technology. Background technique [0002] The informatization of social life and the globalization of economy have made language learning more and more important. As one of the most important information carriers, language has become the most widely used language in various fields of human life. [0003] Taking English as an example, face-to-face classroom teaching is the most widely used form of oral English learning in China. Due to the lack of English teachers, it is difficult for students to obtain a one-to-one English learning environment. In the face of this situation, many units have developed syste...

AI Technical Summary

Problems solved by technology

Among the open-ended question types, oral short dialogue is a common question type. In recent years, some attempts have been made in the field of computer-assisted language learning at home and abroad for open-ended question types. For example, the American Educational Testing Service (ETS) developed a test for SpeechRater is a scoring system for oral composition, but the problem with these systems is that the recognition rate is poor due to the incomplete coverage of expert knowledge, and the content other than expert knowledge cannot be processed (or the effect is very poor). There are no mature products and technical solutions for grading Chinese-style short conversations

Images