Attention-Based Transformer-BiGRU for Question Classification

A question answering (QA) system is a research direction in the field of artificial intelligence and natural language processing (NLP) that has attracted much attention broad development prospects. As one main components QA system, accuracy classification plays key role entire task. Therefore, not only traditional machine learning methods but also today’s deep are widely used deeply studied tasks. This paper mainly introduces our work on two aspects Chinese classification. The first to use an answer-driven method build richer dataset for small-scale problems existing experimental dataset, which certain reference value expansion especially construction those low-resource datasets. second propose model problem with Transformer + Bi-GRU Attention structure. strong coding ability, it adopts scheme fixed length, divides long text into multiple segments, each segment coded separately; there no interaction occurs between segments. Here, we achieve information segments through so as improve effect sentences. Our purpose adding mechanism highlight semantics questions contain answers. results show proposed this significantly improved