We study the problem of entity detection and normalization applied to patient self-reports of symptoms that arise as side-effects of vaccines. Our application domain presents unique challenges that render traditional classification methods ineffective: the number of entity types is large; and many symptoms are rare, resulting in a long-tail distribution of training examples per entity type. We tackle these challenges with an autoregressive model that generates standardized names of symptoms. We introduce a data augmentation technique to increase the number of training examples for rare symptoms. Experiments on real-life patient vaccine symptom self-reports show that our approach outperforms strong baselines, and that additional examples improve performance on the long-tail entities.
Multi-Task Learning for Knowledge Graph Completion with Pre-trained Language Models
Bosung Kim | Taesuk Hong | Youngjoong Ko | Jungyun Seo
Proceedings of the 28th International Conference on Computational Linguistics
As research on utilizing human knowledge in natural language processing has attracted considerable attention in recent years, knowledge graph (KG) completion has come into the spotlight. Recently, a new knowledge graph completion method using a pre-trained language model, such as KG-BERT, is presented and showed high performance. However, its scores in ranking metrics such as Hits@k are still behind state-of-the-art models. We claim that there are two main reasons: 1) failure in sufficiently learning relational information in knowledge graphs, and 2) difficulty in picking out the correct answer from lexically similar candidates. In this paper, we propose an effective multi-task learning method to overcome the limitations of previous works. By combining relation prediction and relevance ranking tasks with our target link prediction, the proposed model can learn more relational properties in KGs and properly perform even when lexical similarity occurs. Experimental results show that we not only largely improve the ranking performances compared to KG-BERT but also achieve the state-of-the-art performances in Mean Rank and Hits@10 on the WN18RR dataset.