While it is already possible to construct a large-scale English corpus easily and inexpensively using web documents, such documents have problems of reliability with respect to their English quality. Thus, we have developed a system that automatically gathers English technical papers from the web, sorts them into those written by native speakers and those written by non-native speakers, and then uses them to construct a native speaker and non-native speaker corpus. We discuss a method of using the corpus for providing alternative candidates for appropriate adjectives against unnatural English adjective-noun co-occurrence constructions 〈a,n〉. The appropriateness of adjective a' is evaluated based on the similarity of the occurring environments between a and a'.
|Number of pages||10|
|Journal||Procedia - Social and Behavioral Sciences|
|Publication status||Published - 2011|
|Event||Conference on Pacific Association for Computational Linguistics, PACLING 2011 - Kuala Lumpur, Malaysia|
Duration: Jul 19 2011 → Jul 21 2011
All Science Journal Classification (ASJC) codes
- Social Sciences(all)