Providing effective feedback on whole-phrase input in computer-assisted language learning

CALL systems which allow whole-phrase input are still in the minority and those which do rely either on parsing for feedback provision (a solution which is ineffective when input is poor) or simply process input token by token (an unsatisfactory method if input is incorrectly ordered). Since poor input and incorrect word-order occur frequently in CALL responses a different approach may be beneficial. The LISC system, developed at the University of Kent to present phrasetranslation exercises, uses an error-detection and feedback mechanism based on fine-granularity sequence comparison. It compares input phrases to acceptable (correct) answers, but unlike traditional error-checking routines it does not fail on encountering unordered or ungrammatical input. An added advantage of its error detection method is that the system is language independent.