Towards Improving Synonym Options in a Text Modification Application

The Language Musesystem (LM) is an application that supports classroom teachers in text modification for middleand highschool learners who are non-native English speakers (NNES). The application performs linguistic analysis on classroom texts highlighting lexical, syntactic, and pragmatic features, indicating potential areas of linguistic complexity. One of the lexical feature options that LM offers are synonym candidates for words in a text. This study investigates how to improve the current synonym detection feature, using a distributional method that identifies similar words [Lin98, LC03] and WordNet. For single words and multi-word expressions, human judges annotated “acceptability” of synonym candidates from both resources. Humans attained high levels of agreement. Outcomes indicated that the distributional method and WordNet provide complementary options that together could provide improved resources.