Manual and Automatic Creation of Learner Corpora

Description

Competence in written English has become essential for effective international communication. To help ESL (English as a Second Language) learners acquire this competence, there has been much effort to develop writing assistance tools. State-of- the-art writing assistance tools are built upon statistical models of non-native writing errors, estimated from large corpora containing typical learner errors. Creating these corpora, however, is time consuming and labor intensive. The goal of this proposal is two-fold: (1) to produce an error-annotated corpus of non-native writing samples for the research community, and (2) to use the corpus to develop and validate computational techniques for automatically creating similar corpora using web data./p>