Dialogue Systems for Computer-Assisted Language Learning (CALL)
Speaker:Sabrina Wilske
Institution: Saarland University
Abstract:
Computer applications for second language learning, ( e.g. [Seneff et al, 2004] ), can fill the gap between the conventional learning situation between teacher and student and the student using static media as books or tapes to improve her language skills. Interactive dialogue systems can be employed as a surrogate for a real language partners or teachers. Simulating the full range of conversational capabilities and language skills of a human being is still out of reach, but an artificial system has its advantages over a human. Learners can deploy it whenever they want, they can focus on specific issues, and the system is more patient and tolerant to repetitions and can provide a less threatening communication situation than speaking to a real person would be [Seneff et al, 2004].
Working on a tutorial dialogue system involves a number of interesting and challenging issues that I would like to address in my research:
- Content: What is the dialogue about? The dialogue should focus on the given learning objectives for a certain lesson and the system's utterances should be constrained to vocabulary and syntactical constructions learned so far. Hence the dialogue should be parameterizable to a subset of the lexicon, grammatical constructions, and possibly a certain topic or domain.
- Initiative: The dialogue should be mixed-initiative, raising the question of when to take initiative and what kind of?
- Errors in user's input should be detected, analyzed, classified and remedied by correcting, explaining and exercising.
- Learner modeling: The system should be adaptable to the user (preferences of any kind, interest in topics, proficiency level, native/background languages, proneness to errors, etc.) [Murphy and McTear 97].
- Repetition rate: When and how often should the system repeat vocabulary/grammar learned at an earlier point of time in order to achieve optimal learning results?
Approaching these problems I will have to make a number of content and design decisions, starting with the target language(s), the modality of interaction (typing or talking), constraints of the dialogue according to topics of different lessons. At the moment I would like to develop a system for learning Chinese as this is one of the languages I learnt myself (and the most challenging one). Given the difficulty of reliable speech recognition, I would like to start with a system that communicates via typing rather than speaking.