KnowGraph@IITK at SemEval-2021 Task 11: Building KnowledgeGraph for NLP Research

Research in Natural Language Processing is making rapid advances, resulting\nin the publication of a large number of research papers. Finding relevant\nresearch papers and their contribution to the domain is a challenging problem.\nIn this paper, we address this challenge via the SemEval 2021 Task 11:\nNLPContributionGraph, by developing a system for a research paper\ncontributions-focused knowledge graph over Natural Language Processing\nliterature. The task is divided into three sub-tasks: extracting contribution\nsentences that show important contributions in the research article, extracting\nphrases from the contribution sentences, and predicting the information units\nin the research article together with triplet formation from the phrases. The\nproposed system is agnostic to the subject domain and can be applied for\nbuilding a knowledge graph for any area. We found that transformer-based\nlanguage models can significantly improve existing techniques and utilized the\nSciBERT-based model. Our first sub-task uses Bidirectional LSTM (BiLSTM)\nstacked on top of SciBERT model layers, while the second sub-task uses\nConditional Random Field (CRF) on top of SciBERT with BiLSTM. The third\nsub-task uses a combined SciBERT based neural approach with heuristics for\ninformation unit prediction and triplet formation from the phrases. Our system\nachieved F1 score of 0.38, 0.63 and 0.76 in end-to-end pipeline testing, phrase\nextraction testing and triplet extraction testing respectively.\n