PETS计算机辅助口试自动评分技术研究  

A study on automated scoring of PETS computer-based speaking test

作　　者：乔辉[1] 董滨[2] 刘常亮[2] 

机构地区：[1]教育部考试中心,北京100084 [2]中国科学院声学研究所,北京100190

出　　处：《外语测试与教学》2012年第3期47-52,共6页Foreign Language Testing and Teaching

基　　金：全国教育科学规划教育部重点课题"计算机辅助英语测试系统的研究与实践"(编号GFA097014)的研究成果之一

摘　　要：本文首先介绍了针对开放式题型的自动评分技术,然后以EduRater评分系统为例阐述如何构建一个实际的考试评分系统,并重点介绍了该研究中遇到的问题及解决方法,如声学模型和识别词表的确定、语言模型和语法模型的应用等。研究人员利用从PETS实考考生中得到的数据,以评分专家所给定的分数作为计算机评分比较的对象,对自动评分系统的信度进行深入研究,证实了该系统的实用性。This paper first explains the automated scoring methods for open-ended questions and then, with the intro- duction of an automated scoring system -- EduRater, it explains how to design a practical scoring system for a speak- ing test. Problems emerging during the research and solutions to them are discussed, such as the selection of acoustic models, the development of recognizable word list, the application of linguistic models and grammatical models. Data obtained from a PETS-I live test are used to study the reliability of the automated scoring system. The practicability of the system is verified by comparing to human experts scoring.

关 键 词：计算机辅助考试 口试 人工评分 自动评分 

分 类 号：H319[语言文字—英语]