Learning Semantic Textual Similarity from Conversations
This work addresses semantic textual similarity for natural language processing applications, but it is incremental as it builds on existing methods with multitask training.
The authors tackled the problem of learning sentence-level semantic similarity by training an unsupervised model on conversational data to predict input-response pairs, achieving the best performance among neural models on the STS benchmark and competitive results on the CQA question similarity subtask.
We present a novel approach to learn representations for sentence-level semantic similarity using conversational data. Our method trains an unsupervised model to predict conversational input-response pairs. The resulting sentence embeddings perform well on the semantic textual similarity (STS) benchmark and SemEval 2017's Community Question Answering (CQA) question similarity subtask. Performance is further improved by introducing multitask training combining the conversational input-response prediction task and a natural language inference task. Extensive experiments show the proposed model achieves the best performance among all neural models on the STS benchmark and is competitive with the state-of-the-art feature engineered and mixed systems in both tasks.