WordNet Gloss Translation for Under-resourced Languages using Multilingual Neural Machine Translation

In this paper, we translate the glosses in the English WordNet based on the expand approach for improving and generating wordnets with the help of multilingual neural machine translation. Neural Machine Translation (NMT) has recently been applied to many tasks in natural language processing, leading to state-of-the-art performance. However, the performance of NMT often suffers in low resource scenarios where large corpora cannot be obtained. Using training data from closely related language have proven to be invaluable for improving performance. In this paper, we describe how we trained multilingual NMT from closely related language utilizing phonetic transcription for Dravidian languages. We report the evaluation result of the generated wordnets sense in terms of precision. By comparing to the recently proposed approach, we show improvement in terms of precision.