Doc2EDAG: An End-to-End Document-level Framework for Chinese Financial Event Extraction

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 337–346, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 337 Doc2EDAG: An End-to-End Document-level Framework for Chinese Financial Event Extraction Shun Zheng†∗, Wei Cao‡, Wei Xu†, Jiang Bian‡ †Institute of Interdisciplinary Information Sciences, Tsinghua University ‡Microsoft Research zhengs14@mails.tsinghua.edu.cn; {Wei.Cao, Jiang.Bian}@microsoft.com; weixu@mail.tsinghua.edu.cn Abstract Most existing event extraction (EE) methods merely extract event arguments within the sen- tence scope. However, such sentence-level EE methods struggle to handle soaring amounts of documents from emerging applications, such as finance, legislation, health, etc., where event arguments always scatter across different sen- tences, and even multiple such event men- tions frequently co-exist in the same docu- ment. To address these challenges, we pro- pose a novel end-to-end model, Doc2EDAG, which can generate an entity-based directed acyclic graph to fulfill the document-level EE (DEE) effectively. Moreover, we reformal- ize a DEE task with the no-trigger-words de- sign to ease document-level event labeling. To demonstrate the effectiveness of Doc2EDAG, we build a large-scale real-world dataset con- sisting of Chinese financial announcements with the challenges mentioned above. Ex- tensive experiments with comprehensive anal- yses illustrate the superiority of Doc2EDAG over state-of-the-art methods. Data and codes can be found at https://github.com/ dolphin-zs/Doc2EDAG. 1 Introduction Event extraction (EE), traditionally modeled as detecting trigger words and extracting correspond- ing arguments from plain text, plays a vital role in natural language processing since it can produce valuable structured information to facilitate a vari- ety of tasks, such as knowledge base construction, question answering, language understanding, etc. In recent years, with the rising trend of digital- ization within various domains, such as finance, legislation, health, etc., EE has become an increas- ingly important accelerator to the development of ∗ This work was done during the internship of Shun Zheng at Microsoft Research Asia, Beijing, China. 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 Year 0 10 20 30 # Announcements (x1000) Figure 1: The rapid growth of event-related announce- ments considered in this paper. business in those domains. Take the financial do- main as an example, continuous economic growth has witnessed exploding volumes of digital finan- cial documents, such as financial announcements in a specific stock market as Figure 1 shows, spec- ified as Chinese financial announcements (ChFi- nAnn). While forming up a gold mine, such large amounts of announcements call EE for assisting people in extracting valuable structured informa- tion to sense emerging risks and find profitable op- portunities timely. Given the necessity of applying EE on the finan- cial domain, the specific characteristics of finan- cial documents as well as those within many other business fields, however, raise two critical chal- lenges to EE, particularly arguments-scattering and multi-event. Specifically, the first challenge indicates that arguments of one event record may scatter across multiple sentences of the docu- ment, while the other one reflects that a doc- ument is likely to …

Content truncated to 500 words