EMNLP2020 | 近期必读Question Answering精选论文

1.论文名称：Hierarchical Graph Network for Multi-hop Question Answering

• In contrast to one-hop question answering, where answers can be derived from a single paragraph, recent studies have more and more focused on multihop reasoning across multiple documents or paragraphs for question answering.

• The authors propose a new approach, Hierarchical Graph Network (HGN), for multi-hop question answering.

• To capture clues from different granularity levels, the HGN model weaves heterogeneous nodes into a single unified graph.

• Experiments with detailed analysis demonstrate the effectiveness of the proposed model, which achieves state-of-the-art performance on HotpotQA benchmark.

2.论文名称：Scalable Multi-Hop Relational Reasoning for Knowledge-Aware Question Answering

• Many recently proposed question answering tasks require machine comprehension of the question and context, and relational reasoning over entities and their relationships based by referencing external knowledge.

• We propose a novel graph encoding architecture, Multi-hop Graph Relation Networks (MHGRN), which combines the strengths of path-based models and graph neural network (GNN).

• The proposed MHGRN generalizes and combines the advantages of GNNs and path-based reasoning models.

• It explicitly performs multi-hop relational reasoning and is empirically shown to outperform existing methods with superior scalablility and interpretability.

3.论文名称：Training Question Answering Models From Synthetic Data

• One of the limitations of developing models for question answering, or any Deep Learning application for that matter, is the availability and cost of labeled training data.

• The authors build upon existing work in large scale language modeling and question generation to push the quality of synthetic question generation.

• Finetuning the resulting model on real SQUAD1.1 data further boosts the EM score to 89.4.

• The authors generate synthetic text from a Wikipedia-finetuned GPT-2 model, generate answer candidates and synthetic questions based on those answers, and train a BERT-Large model to achieve similar question answering accuracy without directly using any real data at all.

4.论文名称：Look at the First Sentence: Position Bias in Question Answering

• Most QA studies frequently utilize start and end positions of answers as training targets without much considerations.

• The authors' study shows that most QA models fail to generalize over different positions when trained on datasets having answers in a specific position.

• The authors introduce several de-biasing methods to make models to ignore the spurious positional cues, and find out that the sentence-level answer prior is very useful.

• The authors' findings generalize to different positions and different datasets.

5.论文名称：Infusing Disease Knowledge into BERT for Health Question Answering, Medical Inference, Disease Name Recognition

• Human disease is “a disorder of structure or function in a human that produces specific signs or symptoms”.

• The authors propose a new disease infusion training procedure to augment BERT-like pre-trained language models with disease knowledge.

• The authors conduct this training procedure on a suite of BERT models and evaluate them over disease related tasks.

• Experimental results show that these models can be enhanced by this disease infusion method in most cases.

6.论文名称：Self-supervised Knowledge Triplet Learning for Zero-shot Question Answering

• The ability to understand natural language and answer questions is one of the core focus in the field of natural language processing.

• The authors achieve state-of-the-art results for zero-shot and propose a strong baseline for the few-shot question answering task.

• The authors can see the Transformer encoder trained on KTL perform significantly better than the baseline models in this setting.

• The authors' framework achieves state-of-the-art in the zero-shot question answering task and sets a strong baseline in the few-shot question answering task.