Evidence retrieval is a critical stage of question answering (QA), necessary not only to improve performance, but also to explain the decisions of the QA method. We introduce a simple, fast, and unsupervised iterative evidence retrieval method, which relies on three ideas: (a) an unsupervised alignment approach to soft-align questions and answers with justification sentences using only GloVe embeddings, (b) an iterative process that reformulates queries focusing on terms that are not covered by existing justifications, which (c) stops when the terms in the given question and candidate answers are covered by the retrieved justifications. Despite its simplicity, our approach outperforms all the previous methods (including supervised methods) on the evidence selection task on two datasets: MultiRC and QASC. When these evidence sentences are fed into a RoBERTa answer classification component, we achieve state-of-the-art QA performance on these two datasets.

Large pretrained language models like BERT, after fine-tuning to a downstream task, have achieved high performance on a variety of NLP problems. Yet explaining their decisions is difficult despite recent work probing their internal representations. We propose a procedure and analysis methods that take a hypothesis of how a transformer-based model might encode a linguistic phenomenon, and test the validity of that hypothesis based on a comparison between knowledge-related downstream tasks with downstream control tasks, and measurement of cross-dataset consistency. We apply this methodology to test BERT and RoBERTa on a hypothesis that some attention heads will consistently attend from a word in negation scope to the negation cue. We find that after fine-tuning BERT and RoBERTa on a negation scope task, the average attention head improves its sensitivity to negation and its attention consistency across negation datasets compared to the pre-trained models. However, only the base models (not the large models) improve compared to a control task, indicating there is evidence for a shallow encoding of negation only in the base models.

Concept normalization, the task of linking textual mentions of concepts to concepts in an ontology, is challenging because ontologies are large. In most cases, annotated datasets cover only a small sample of the concepts, yet concept normalizers are expected to predict all concepts in the ontology. In this paper, we propose an architecture consisting of a candidate generator and a list-wise ranker based on BERT. The ranker considers pairings of concept mentions and candidate concepts, allowing it to make predictions for any concept, not just those seen during training. We further enhance this list-wise approach with a semantic type regularizer that allows the model to incorporate semantic type information from the ontology during training. Our proposed concept normalization framework achieves state-of-the-art performance on multiple datasets.

In undergraduate theses, a good methodology section should describe the series of steps that were followed in performing the research. To assist students in this task, we develop machine-learning models and an app that uses them to provide feedback while students write. We construct an annotated corpus that identifies sentences representing methodological steps and labels when a methodology contains a logical sequence of such steps. We train machine-learning models based on language modeling and lexical features that can identify sentences representing methodological steps with 0.939 f-measure, and identify methodology sections containing a logical sequence of steps with an accuracy of 87%. We incorporate these models into a Microsoft Office Add-in, and show that students who improved their methodologies according to the model feedback received better grades on their methodologies.

Recently BERT has achieved a state-of-the-art performance in temporal relation extraction from clinical Electronic Medical Records text. However, the current approach is inefficient as it requires multiple passes through each input sequence. We extend a recently-proposed one-pass model for relation classification to a one-pass model for relation extraction. We augment this framework by introducing global embeddings to help with long-distance relation inference, and by multi-task learning to increase model performance and generalizability. Our proposed model produces results on par with the state-of-the-art in temporal relation extraction on the THYME corpus and is much “greener” in computational cost.