GRIT: Generative Role-filler Transformers for Document-level Event Entity Extraction

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 634–644 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 634 GRIT: Generative Role-filler Transformers for Document-level Event Entity Extraction Xinya Du Alexander M. Rush Department of Computer ScienceClaire Cardie Cornell University {xdu, cardie}@cs.cornell.edu arush@cornell.edu Abstract We revisit the classic problem of document- level role-filler entity extraction (REE) for tem- plate filling. We argue that sentence-level ap- proaches are ill-suited to the task and intro- duce a generative transformer-based encoder- decoder framework (GRIT) that is designed to model context at the document level: it can make extraction decisions across sentence boundaries; is implicitly aware of noun phrase coreference structure, and has the capacity to respect cross-role dependencies in the tem- plate structure. We evaluate our approach on the MUC-4 dataset, and show that our model performs substantially better than prior work. We also show that our modeling choices con- tribute to model performance, e.g., by implic- itly capturing linguistic knowledge such as rec- ognizing coreferent entity mentions. 1 Introduction Document-level template filling (Sundheim, 1991, 1993; Grishman and Sundheim, 1996) is a clas- sic problem in information extraction (IE) and NLP (Jurafsky and Martin, 2014). It is of great importance for automating many real-world tasks, such as event extraction from newswire (Sundheim, 1991). The complete task is generally tackled in two steps. The first step detects events in the article and assigns templates to each of them (template recognition); the second step performs role-filler entity extraction (REE) for filling in the templates. In this work we focus on the role-filler entity ex- traction (REE) sub-task of template filling (Fig- ure 1).1 The input text describes a bombing event; the goal is to identify the entities that fill any of the roles associated with the event (e.g., the perpetra- tor, their organization, the weapon) by extracting 1In this work, we assume there is one generic template for the entire document (Huang and Riloff, 2011, 2012). Role Role-filler Entities Perpetrator Individual two men, two men wearing sports clothes, Shining Path members Perpetrator Organization Shining Path Physical Target water pipes, water pipes Pilmai telephone company building, telephone company building, telephone company offices public telephone booth Weapon 125 to 150 grams of TnT Victim - Gold extractions: Input document: … A bomb exploded in a Pilmai alley destroying some [water pipes]. According to unofficial reports, the bomb contained [125 to 150 grams of TnT] and was placed in the back of the [Pilmai [telephone company building]]. The explosion occurred at 2350 on 16 January, causing panic but no casualties. The explosion caused damages to the [telephone company offices]. It also destroyed a [public telephone booth] and [water pipes]. Witnesses reported that the bomb was planted by [[two men] wearing sports clothes], who escaped into the night. … They were later identified as [[Shining Path] members]. Figure 1: Role-filler entity extraction (REE). The first mention of each role-filler entity is bold in the table and document. The arrows denote …

Content truncated to 500 words