Joint Training of Candidate Extraction and Answer Selection for Reading Comprehension

By Zhen Wang, Xinyan Xiao, Yajuan Lyu, Tian Wu. 2018

38 citations recorded by the source corpus.

While sophisticated neural-based techniques have been developed in reading comprehension, most approaches model the answer in an independent manner, ignoring its relations with other answer candidates. This problem can be even worse in open-domain scenarios, where candidates from multiple passages should be combined to answer a single question. In this paper, we formulate reading comprehension as an extract-then-select two-stage procedure. We first extract answer candidates from passages, then select the final answer by combining information from all the candidates. Furthermore, we regard candidate extraction as a latent variable and train the two-stage process jointly with reinforcement learning. As a result, our approach has improved the state-of-the-art performance significantly on two challenging open-domain reading comprehension datasets. Further analysis demonstrates the effectiveness of our model components, especially the information fusion of all the candidates and the joint training of the extract-then-select procedure.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Question Answering & Reading Comprehension

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Question Answering & Reading Comprehension

Citations in the corpus

Retrieve, Read, Rerank: Towards End-to-End Multi-Document Reading Comprehension (2019, unknown): “, 2018a) 3 3 7 Pipeline* Extract-Select (Wang et al., 2018d) 7 3 3 Pipeline* V-Net (Wang et al.”
Neural Arabic Question Answering (2019, background): “Other approaches have attempted to build more sophisticated retrievers by formulating it as a reinforcement learning problem (Wang et al., 2018b,a), or as a supervised learning problem us-ing distant supervision for data (Das et al., 2018; Lin et al., 2018).”
Exploring Graph-structured Passage Representation for Multi-hop Reading Comprehension with Graph Neural Networks (2018, methodology): “Previous work (Wang et al., 2018a) shows that summing the probabilities over all occurrences of the same entity mention is important for the multi-passage scenario. α k is the attention score for the entity mention (cid:15) k , calculated by an additive attention model shown below: where v a , W a…”