|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Home | Research | Research Programs | Information Analysis and Retrieval |
Information Analysis and RetrievalAdverbs: Intelligent Computer Aided Language Learning (iCALL)Most CALL systems are limited by their language examples, especially if they are designed to provide students with meaningful positive and negative feedback. Generally, native speakers must examine the language data used and possible responses by students to determine acceptability. The cost of doing this means that such systems have few examples and only limited flexibility. An obvious alternative is to let the computer decide whether a language example or a student's response is acceptable. For example, if students are asked to type a sentence using an adverb, an intelligent system could determine whether the adverb has been inserted correctly and, if not, indicate its proper position in the sentence. Such intelligent systems require Natural Language Processing (NLP), such as Nagata 2002. NRC-IIT research has led to a statistical NLP technique based on maximum entropy. The technique estimates the probability of a particular adverb being placed in a specific position within a sentence. Statistics are collected over a large corpus of language instances in actual use. NRC-IIT researchers expect that this technique will apply to many other aspects of syntax and semantics and in any language. The NRC-IIT technique allows for much more attractive and flexible CALL systems. The technique permits an almost unlimited number of language examples, because they do not have to be examined by native speakers. With this technique, students can enter their own input sentences with more flexibility than is possible in other NLP or CALL systems. Research ContactDr. Joel Martin Business ContactRandall Milburn |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|