Recent work has demonstrated that entity representations can be extracted from pre-trained language models to develop knowledge graph completion models that are more robust to the naturally occurring sparsity found in knowledge graphs. In this work, we conduct a comprehensive exploration of how to best extract and incorporate those embeddings into knowledge graph completion models. We explore the suitability of the extracted embeddings for direct use in entity ranking and introduce both unsupervised and supervised processing methods that can lead to improved downstream performance. We then introduce supervised embedding extraction methods that can extract more informative representations. We then synthesize our findings and develop a knowledge graph completion model that significantly outperforms recent neural models.

Knowledge Graph (KG) completion research usually focuses on densely connected benchmark datasets that are not representative of real KGs. We curate two KG datasets that include biomedical and encyclopedic knowledge and use an existing commonsense KG dataset to explore KG completion in the more realistic setting where dense connectivity is not guaranteed. We develop a deep convolutional network that utilizes textual entity representations and demonstrate that our model outperforms recent KG completion methods in this challenging setting. We find that our model’s performance improvements stem primarily from its robustness to sparsity. We then distill the knowledge from the convolutional network into a student network that re-ranks promising candidate entities. This re-ranking stage leads to further improvements in performance and demonstrates the effectiveness of entity re-ranking for KG completion.

In this paper, we discuss our submission for DialDoc subtask 1. The subtask requires systems to extract knowledge from FAQ-type documents vital to reply to a user’s query in a conversational setting. We experiment with pretraining a BERT-based question-answering model on different QA datasets from MRQA, as well as conversational QA datasets like CoQA and QuAC. Our results show that models pretrained on CoQA and QuAC perform better than their counterparts that are pretrained on MRQA datasets. Our results also indicate that adding more pretraining data does not necessarily result in improved performance. Our final model, which is an ensemble of AlBERT-XL pretrained on CoQA and QuAC independently, with the chosen answer having the highest average probability score, achieves an F1-Score of 70.9% on the official test-set.

Automated radiology report generation has the potential to reduce the time clinicians spend manually reviewing radiographs and streamline clinical care. However, past work has shown that typical abstractive methods tend to produce fluent, but clinically incorrect radiology reports. In this work, we develop a radiology report generation model utilizing the transformer architecture that produces superior reports as measured by both standard language generation and clinical coherence metrics compared to competitive baselines. We then develop a method to differentiably extract clinical information from generated reports and utilize this differentiability to fine-tune our model to produce more clinically coherent reports.