A POST Parser-Based Learner Model for Template-Based ICALL for Japanese-English Writing Skills

Liang Chen, Naoyuki Tokuda, Dahai Xiao
2002 Computer Assisted Language Learning  
Improving the parsing accuracy of a statistical 'context sensitive' parser, we have developed a new learner model which is capable of identifying structural deficiencies of a learner for the template-automaton-based ICALL (intelligent language tutoring system) of Japanese-English writing skill [1]. The parsing accuracy of corpus-based parsers is improved first by pre-assigning POS (part-of-speech) tags to well-formed sentences in a template and then by making use of a compound dictionary of
more » ... d dictionary of idiomatic phrases. By defining the minimum unit of a parsed tree as a Minimum Error Sub-Tree if the unit contains a HCS (heaviest common sequence)-based static error(s) identified, we show how the new syntactic-based learner model can now be used to identify structural deficiencies of a learner largely due to structural differences between his first and second languages, thus introducing an entirely new scheme of remediation in the ICALL system.
doi:10.1076/call.15.4.357.8267 fatcat:qbdpioourvdxfgomsbpqqxlrfe