NegBERT: A Transfer Learning Approach for Negation Detection and Scope Resolution

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 5739–5748 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 5739 NegBERT: A Transfer Learning Approach for Negation Detection and Scope Resolution Aditya Khandelwal, Suraj Sawant College of Engineering Pune khandelwalar16.comp@coep.ac.in, sts.comp@coep.ac.in Abstract Negation is an important characteristic of language, and a major component of information extraction from text. This subtask is of considerable importance to the biomedical domain. Over the years, multiple approaches have been explored to address this problem: Rule-based systems, Machine Learning classifiers, Conditional Random Field models, CNNs and more recently BiLSTMs. In this paper, we look at applying Transfer Learning to this problem. First, we extensively review previous literature addressing Negation Detection and Scope Resolution across the 3 datasets that have gained popularity over the years: the BioScope Corpus, the Sherlock dataset, and the SFU Review Corpus. We then explore the decision choices involved with using BERT, a popular transfer learning model, for this task, and report state-of-the-art results for scope resolution across all 3 datasets. Our model, referred to as NegBERT, achieves a token level F1 score on scope resolution of 92.36 on the Sherlock dataset, 95.68 on the BioScope Abstracts subcorpus, 91.24 on the BioScope Full Papers subcorpus, 90.95 on the SFU Review Corpus, outperforming the previous state-of-the-art systems by a significant margin. We also analyze the model’s generalizability to datasets on which it is not trained. Keywords: Negation, Scope Resolution, Transfer Learning 1. Introduction Negation Detection and Scope Resolution is an impor- tant subtask for tasks ranging from Sentiment Analysis, where the sentiment of a given sentence is dependent on negation, to query response systems like Chatbots, where negation entirely changes the meaning and hence the rel- evance of a certain body of text. A substantial portion of the research till date on this topic focused solely on data from the biomedical domain, where use of negation cues is abundant (eg. medical reports). While negation is intu- itive for humans to spot, finding the exact words that indi- cate such negation and delineating the scope of such nega- tion cues has proven to be a tricky problem for computer- based systems. One could imagine that finding negation cues and their scopes could be easily solved via rules and carefully designed heuristics, and this was the exact ap- proach used by the initial few systems attempting this task. But given the complexities of human language, these ap- proaches weren’t accurate enough. Thus, other methods were explored, and Deep Learning-based approaches have shown to be particularly promising. A simple example of negation is as follows: This is not [a negation]. We can observe that ‘not’ is the negation word (known as the negation cue) and the words whose meaning is al- tered by ‘not’ are ‘a’ and ‘negation’, which belong to what is known as the cue’s scope. Negation detection involves finding these negation cues, and scope resolution for each cue necessitates finding the words affected negatively by …

Content truncated to 500 words