Kathleen McCoy (U. of Delaware):
Generation Issues in a Second Language Learning Tool
ABSTRACT Traditionally the task of natural language generation has been broken into two components: what to say (i.e., what information should be included in a response and how should it be structured) and how to say it (i.e., what syntactic structures should be chosen for each generated sentence). In this talk I discuss some issues that these two components must take into account in the context of a writing tool for teaching English as a second language. Our long-term goal is to develop a computer-assisted language learning (CALL) tool to help deaf students learn written English. The targeted students are users of American Sign Language (ASL), a language that is very different from English in its structure and discourse strategies. The approach we take is to view the student's learning of written English as a task in second language acquisition. I argue that in order for the system to be an effective language tutor, it must consider the generation process of the student. By this I mean that the system must be aware of what English constructions the student has already mastered, what constructions the student is in the process of mastering, and what constructions are beyond the student's current ability. I discuss how this information affects the generation decisions of the tutoring system and describe our method for capturing how this information changes over time. The models are motivated through describing its application in the project geared for native users of American Sign Language (ASL) learning written English, but I indicate how the models might be used by learners coming from other native languages as well.
BIOGRAPHY I am an associate professor in the Dept. of Computer and Information Sciences at the University of Delaware which I joined in September of 1985. I received my PhD in the Dept. of Computer and Information Science of the University of Pennsylvania 1985, and my MS from the same institution in 1982. I have worked in the general area of natural language generation since my graduate student days and have been an active member of that and the wider computational linguistics community. For example, I have served on the editorial board of the Computational Linguistics Journal and have served on the executive committee of the Association for Computational Linguistics. My work has spanned additional areas including natural language discourse (particularly the areas of focus and pronoun resolution), user modeling, AI and education, intelligent user interfaces, and interfaces designed to enhance communicative abilities of people with disabilities.