Abstract

Inspired by conversational reading comprehension (CRC), this paper studies anovel task of leveraging reviews as a source to build an agent that can answermulti-turn questions from potential consumers of online businesses. We firstbuild a review CRC dataset and then propose a novel task-aware pre-tuning steprunning between language model (e.g., BERT) pre-training and domain-specificfine-tuning. The proposed pre-tuning requires no data annotation, but cangreatly enhance the performance on our end task. Experimental results show thatthe proposed approach is highly effective and has competitive performance asthe supervised approach. The dataset is available at\url{https://github.com/howardhsu/RCRC}

 

Quick Read (beta)