Abstract:
Computers have played a significant role in second language learning as they can offer different types of language activities and provide learners with immediate and appropriate feedbacks. Besides that, learning is activated by focusing the learner's attention to the correct form and comparing it to the wrong one.
The focus of this dissertation is set on GI-Tutor, a German grammar checker designed for Italian native speakers learning German. The system's structure was taken from a previous study and enforced to analyse a conspicuous number of sentences, with different phrase structures. The lexicon used has been obtained through an adaptation of the lexicon made available by Hamburg University Constraint Dependency Grammar (JWCDG) in their website. A corpus containing wrong sentences was expanded by extracting data from exams written by students in the first year of the German course at the University Ca' Foscari. The errors were then classified in order to obtain a general statistical analysis of the main problems in the acquisition of German.
Another focus was set on parsers and their use and functionality in language learning.
Furthermore, the performance of the constituency grammar checker was evaluated to determine the types and frequencies of errors it can successfully diagnose. This was done by comparing it to ParZu - a generic German dependency parser developed at the University Zürich.
