Linggle 2.0: a collocation retrieval system with quality example sentences

Linggle is a pattern-based referencing tool that assists in collocation learning. In this ongoing project, we aimed to improve its performance further. First, many of the example sentences are long and difficult for students to understand, so we used a machine learning method and trained a classifier to help select dictionary-like example sentences. Second, we created a database of 60,270,000 sentences from 4C, S2ORC, and VOA Learning English. We also included Google books for real-time supplements. Then, we applied the classifier to select good example sentences from the database for display. We also limited the number of example sentences displayed for search results to improve users’ experiences. Two classes of English as a Foreign Language (EFL) college students (N=51) were invited to use the enhanced tool and filled out a questionnaire. The results showed that the students were positive about Linggle’s new interface and the quality of the example sentences. We expect that more EFL learners will benefit from the tool.