Intelligent computer-assisted language learning: A bibliography

This bibliography focuses on works discussing intelligent computer-assisted language learning (ICALL). It includes over 200 entries divided into three sections: (1) collections, special issues and bibliographies, (2) general/theoretical works, (3) specific applications.Alan Bailin (Ph.D., McGill) is Associate Editor ofComputers and the Humanities. He has participated in the field of Intelligent Computer-Assisted Language Learning as a writer, editor, and researcher. He has also published in the areas of linguistics and semiotics.     

浅谈中学计算机辅助教学中存在的问题

在当代中学校利用计算机进行教学的过程中广泛存在着一些问题,该文将这些问题进行列举和阐述。     

Developing and evaluating language courseware

The paper sets out twenty proposals for the development and evaluation of Computer Assisted Language Learning (CALL) programs. These proposals emerge from special characteristics of language instruction and of the use of computers to assist in language instruction. We combine theoretically-based assumptions with empirical findings drawn from investigation of language courseware for Hebrew speakers in Israel. We first list four unique features of language instruction: (1) the object-language-meta-language distinction; (2) computer as written medium vs. language as primary spoken medium; (3) teaching of second language skills vs. linguistics; (4) the computer as an electronic tool vs. the computer as a cognitive entity simulating the speaker. We then show how these unique characteristics of language instruction (mother-tongue and foreign language) impose special proposals on language courseware. These proposals should be observed in the development of language courseware and in the evaluation of such programs. Clearly, these proposals integrate with general courseware proposals. Michal Ephratt (Ph.D., computational linguistics) completed post-doctoral studies at the University of Rochester. She has been on the staff of the Dept. of Hebrew Linguistics, University of Haifa, since 1988. Some of her publications include Root-Pattern Array: The Main Tool of Hebrew Word Formation (Hebrew University, 1985); and What's in a Joke? in Advances in Artificial Intelligence, Natural Language and Knowledge-based Systems (Springer-Verlag, 1990).This paper is based on work the author did as a consultant in the National Courseware Evaluation Department of the Ministry of Education. I wish to thank Esther Diamant, head of the department, for making the study possible.     

基于GMM-UBM和GLDS-SVM的英文发音错误检测方法

将语种和说话人识别的方法应用到英语发音错误检测系统, 提出一种基于广义线性区分序列支持向量机 (Generalized linear discriminant sequence based SVM, GLDS-SVM)的发音错误检测方法. 主要创新点为: 1)提出一种基于状态拼接的特征规整方案, 增强SVM对发音特征的建模能力; 2)提出一种基于多模型融合的模型训练策略, 该策略可以更加充分地利用训练数据, 并在一定程度上解决了由于真实发音错误数据缺乏造成的正负样本不均衡的问题; 3)将GLDS-SVM与基于通用背景模型GMM (Universal background models based GMM, GMM-UBM)的方法进行融合, 以进一步提高发音检错性能. GLDS-SVM和GMM-UBM的融合系统在仿真测试集和真实测试集上的等错误率 (Equal error rate, EER)分别达到9.92%和16.35%. 同时, GLDS-SVM在模型占用空间和运算速度方面均比传统径向基函数 (Radial basic function, RBF)核方法具有明显优势.     

基于Browser/Server的课件系统的设计与实现

讨论基于Browser/Server的智能计算机辅助教学(ICAI)的实现技术,据此采用VB、SQLServer、IIS等开发平台,建立了一个用等数学教学的ICAI试验系统。     

基于JSM和MLP改进发音错误检测的方法

针对发音错误检测的发音字典生成提出基于联合序列多阶模型(Joint-sequence multi-gram, JSM)和多层神经感知(Multi-layer perception, MLP)的方法. 首先使用JSM模型对发音错误进行建模, 将标准发音和错误发音组合为发音对, 表示它们之间的对应关系, 再使用N元文法来统计各发音对之间的关系, 描述错误发音对上下文关系的依赖. 最后使用MLP对发音对之间的关系进行重新建模, 以学习到在相似的上下文条件下发生的相似的错误. 实验证明使用MLP对高阶模型进行概率重估能有效的平滑概率空间, 提高了发音错误检测的性能.     

Computerized comprehension training for whom and under which conditions is it efficient?

The purpose of this study was to examine the differential impact of a computer‐based comprehension training programme according to second graders' comprehender profiles. These profiles differed in the children's abilities to process three types of textual information: literal information, text‐connecting inferences and gap‐filling inferences. The results showed that all the profiles improved significantly after the training on the question type that had caused them the greatest amount of difficulty on the pretest. In addition, factors that could a priori predict the effectiveness of the programme were investigated by comparing children who benefited from the instruction (responders) with those who did not benefit (nonresponders). Before instruction, these children differed in their lexical knowledge level and logical reasoning skills, but not in their memory or comprehension‐monitoring abilities. The results support the claim that children with comprehension difficulties form a heterogeneous group and emphasize the need to examine the differential impact of comprehension instruction according to the characteristics of poor comprehenders.