Neural models have achieved great success on machine reading comprehension (MRC), many of which typically consist of two components: an evidence extractor and an answer predictor.The former seeks the most relevant information from a reference text, while the latter is to locate or generate answers from the extracted evidence.Despite the importance of evidence labels for training the evidence extractor, they are not cheaply accessible, particularly in many non-extractive MRC tasks such as YES/NO question answering and multi-choice MRC.To address this problem, we present a Self-Training method (STM), which supervises the evidence extractor with auto-generated evidence labels in an iterative process.At each iteration, a base MRC model is trained with golden answers and noisy evidence labels.The trained model will predict pseudo evidence labels as extra supervision in the next iteration.We evaluate STM on seven datasets over three MRC tasks.Experimental results demonstrate the improvement on existing MRC models, and we also analyze how and why such a self-training method works in MRC.

In spite of the great advancements of machine reading comprehension (RC), existing RC models are still vulnerable and not robust to different types of adversarial examples. Neural models over-confidently predict wrong answers to semantic different adversarial examples, while over-sensitively predict wrong answers to semantic equivalent adversarial examples. Existing methods which improve the robustness of such neural models merely mitigate one of the two issues but ignore the other. In this article, we address the over-confidence issue and the over-sensitivity issue existing in current RC models simultaneously with the help of external linguistic knowledge. We first incorporate external knowledge to impose different linguistic constraints (entity constraint, lexical constraint, and predicate constraint), and then regularize RC models through posterior regularization. Linguistic constraints induce more reasonable predictions for both semantic different and semantic equivalent adversarial examples, and posterior regularization provides an effective mechanism to incorporate these constraints. Our method can be applied to any existing neural RC models including state-of-the-art BERT models. Extensive experiments show that our method remarkably improves the robustness of base RC models, and is better to cope with these two issues simultaneously.

The ability of story comprehension is a strong indicator of natural language understanding. Recently, Story Cloze Test has been introduced as a new task of machine reading comprehension, i.e., selecting a correct ending from two candidate endings given a four-sentence story context. Most existing methods for Story Cloze Test are essentially matching-based that operate by comparing an individual ending with a given context, therefore suffering from the evidence bias issue: both candidate endings can obtain supporting evidence from the story context, which misleads the classifier to choose an incorrect ending. To address this issue, we present a novel idea to improve story comprehension by utilizing the hints that are obtained through comparing two candidate endings. The proposed model firstly anticipates a feature vector for a possible ending solely based on the context, and then refines the feature prediction using the hints which encode the difference between two candidates. The candidate ending whose feature vector is more similar to the predicted ending vector is regarded as correct. Experimental results demonstrate that our approach can alleviate the evidence bias issue and improve story comprehension.

Multi-relation Question Answering is a challenging task, due to the requirement of elaborated analysis on questions and reasoning over multiple fact triples in knowledge base. In this paper, we present a novel model called Interpretable Reasoning Network that employs an interpretable, hop-by-hop reasoning process for question answering. The model dynamically decides which part of an input question should be analyzed at each hop; predicts a relation that corresponds to the current parsed results; utilizes the predicted relation to update the question representation and the state of the reasoning process; and then drives the next-hop reasoning. Experiments show that our model yields state-of-the-art results on two datasets. More interestingly, the model can offer traceable and observable intermediate predictions for reasoning analysis and failure diagnosis, thereby allowing manual manipulation in predicting the final answer.