CitationIE: Leveraging the Citation Graph for Scientific Information Extraction

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 719–731 August 1–6, 2021. ©2021 Association for Computational Linguistics 719 CitationIE: Leveraging the Citation Graph for Scientific Information Extraction Vijay Viswanathan, Graham Neubig, Pengfei Liu Language Technologies Institute, Carnegie Mellon University {vijayv, gneubig, pfliu3}@cs.cmu.edu Abstract Automatically extracting key information from scientific documents has the potential to help scientists work more efficiently and ac- celerate the pace of scientific progress. Prior work has considered extracting document- level entity clusters and relations end-to-end from raw scientific text, which can improve lit- erature search and help identify methods and materials for a given problem. Despite the im- portance of this task, most existing works on scientific information extraction (SciIE) con- sider extraction solely based on the content of an individual paper, without considering the paper’s place in the broader literature. In con- trast to prior work, we augment our text rep- resentations by leveraging a complementary source of document context: the citation graph of referential links between citing and cited pa- pers. On a test set of English-language scien- tific documents, we show that simple ways of utilizing the structure and content of the cita- tion graph can each lead to significant gains in different scientific information extraction tasks. When these tasks are combined, we ob- serve a sizable improvement in end-to-end in- formation extraction over the state-of-the-art, suggesting the potential for future work along this direction. We release software tools to fa- cilitate citation-aware SciIE development.1 1 Introduction The rapid expansion in published scientific knowl- edge has enormous potential for good, if it can only be harnessed correctly. For example, during the first five months of the global COVID-19 pan- demic, at least 11000 papers were published online about the novel disease (Hallenbeck, 2020), with each representing a potential faster end to a global pandemic and saved lives. Despite the value of this quantity of focused research, it is infeasible 1https://github.com/viswavi/ScigraphIE Speech Papers Vision Papers NLP Papers Citation Graph ML Papers “ [...] The very deep convolutional networks are inspired by the “VGGNet” architecture introduced in [16] for the 2014 ImageNet classification challenge, with the central idea to replace large convolutional kernels by small 3×3 kernels. […] Given the recent popularity of LSTMs for acoustic modeling, we have experimented with such models on the Switchboard task […] ” “The IBM 2016 English Conversational Telephone Speech Recognition System” Interspeech 2016 Salient? No Salient? Yes Figure 1: Example of using the citation graph to im- prove the task of salient entity classification (Jain et al., 2020). In this task, each entity in the document is clas- sified as salient or not, where a salient entity is defined as being relevant to its paper’s main ideas. for the scientific community to read this many pa- pers in a time-critical situation, and make accurate judgements to help separate signal from the noise. To this end, how can machines help researchers quickly identify relevant …

Content truncated to 500 words