Hashing based Answer Selection
This work addresses a practical bottleneck in deploying complex models like BERT for online question answering, offering a memory-efficient solution for real-time applications.
The paper tackles the computational inefficiency of deep models in answer selection by proposing a hashing-based method that learns binary matrix representations for answers, reducing memory cost while maintaining fast online prediction. Experimental results on three datasets show that HAS outperforms existing models to achieve state-of-the-art performance.
Answer selection is an important subtask of question answering (QA), where deep models usually achieve better performance. Most deep models adopt question-answer interaction mechanisms, such as attention, to get vector representations for answers. When these interaction based deep models are deployed for online prediction, the representations of all answers need to be recalculated for each question. This procedure is time-consuming for deep models with complex encoders like BERT which usually have better accuracy than simple encoders. One possible solution is to store the matrix representation (encoder output) of each answer in memory to avoid recalculation. But this will bring large memory cost. In this paper, we propose a novel method, called hashing based answer selection (HAS), to tackle this problem. HAS adopts a hashing strategy to learn a binary matrix representation for each answer, which can dramatically reduce the memory cost for storing the matrix representations of answers. Hence, HAS can adopt complex encoders like BERT in the model, but the online prediction of HAS is still fast with a low memory cost. Experimental results on three popular answer selection datasets show that HAS can outperform existing models to achieve state-of-the-art performance.