Biomedical Event Extraction as Sequence Labeling

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 5357–5367, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 5357 Biomedical Event Extraction as Sequence Labeling Alan Ramponi♦♣ Rob van der Goot♠ Rosario Lombardo♣ Barbara Plank♠ ♦Department of Information Engineering and Computer Science, University of Trento, Italy ♠Department of Computer Science, IT University of Copenhagen, Denmark ♣Fondazione the Microsoft Research – University of Trento Centre for Computational and Systems Biology (COSBI), Italy ramponi@cosbi.eu, robv@itu.dk, lombardo@cosbi.eu, bapl@itu.dk Abstract We introduce Biomedical Event Extraction as Sequence Labeling (BEESL), a joint end- to-end neural information extraction model. BEESL recasts the task as sequence labeling, taking advantage of a multi-label aware encod- ing strategy and jointly modeling the interme- diate tasks via multi-task learning. BEESL is fast, accurate, end-to-end, and unlike current methods does not require any external knowl- edge base or preprocessing tools. BEESL out- performs the current best system (Li et al., 2019) on the Genia 2011 benchmark by 1.57% absolute F1 score reaching 60.22% F1, es- tablishing a new state of the art for the task. Importantly, we also provide first results on biomedical event extraction without gold en- tity information. Empirical results show that BEESL’s speed and accuracy makes it a viable approach for large-scale real-world scenarios.1 1 Introduction Biomedical event extraction provides invaluable means for assisting domain experts in the cura- tion of knowledge bases and biomolecular path- ways (Ananiadou et al., 2010). While the task has received significant attention in research over the last decade, it remains challenging. Progress has been rather stagnating (see Figure 1). Events are typically highly complex and nested structures, which require deep contextual knowl- edge to resolve. This is particularly the case for biomedical NLP (Kim et al., 2011), where biomolecular events can be nested (Miwa et al., 2014) and long-distance arguments are frequent (Li et al., 2019). Figure 2 shows an example with four events. Each event consists of an event mention (trigger) and one or more arguments. For instance, there is a +REGULATION event triggered by the 1The source code is available at https://github. com/cosbi-research/beesl. 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 56 57 58 59 60 F1 score Combination (Joint+parsing) (Riedel et al., 2011) SVM pipeline (+CR) (Miwa et al., 2013) SVM pipeline & MLN (Joint) (Venugopal et al., 2014) Stacked generalization (Majumder et al., 2016) CNN pipeline (+DP) (Bj¨orne & Salakoski, 2018) ensemble single KB-TreeLSTM pipeline (Li et al., 2019) TreeLSTM pipeline BEESL Figure 1: Performance of biomedical event extraction on the BioNLP Genia 2011 test set over time. span “induced”, with a PROTEIN entity (i.e., “IL- 12”) as CAUSE and a nested +REGULATION event (i.e., “activation”) as THEME. Many state-of-the- art biomedical event extraction systems still work as a pipeline and extract event triggers and their arguments independently (Bj¨orne and Salakoski, 2018; Li et al., 2019). They typically employ de- pendency parsing as features in a CNN model en- semble (Bj¨orne and Salakoski, 2018) or in Tree- LSTMs with knowledge bases (Li …

Content truncated to 500 words