Contextual and Semantic Fusion Network for Multiple-Choice Reading Comprehension

Multiple-choice reading comprehension (MCRC) aims to build an intelligent system that automatically selects answer from a candidate set when given passage and question. Existing MCRC systems rarely consider incorporating external knowledge such as explicit semantic information. In this work, we propose Contextual Semantic Fusion Network (CSFN) which effectively integrates contextual representation. CSFN introduces structured semantics pre-trained role labeling. Specially, regard representation important feature fuse with representation, enriches the of sentences. By combining transfer learning strategy, model has better generalization over limited datasets. To evaluate ability our model, conduct experiments on three benchmark datasets: RACE, DREAM, MCTest. Experimental results demonstrate effectiveness proposed model.