Probing into the Root: A Dataset for Reason Extraction of Structural Events from Financial Documents

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 2042–2048 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 2042 Probing into the Root: A Dataset for Reason Extraction of Structural Events from Financial Documents Pei Chen1∗, Kang Liu2,3, Yubo Chen2,3, Taifeng Wang4, and Jun Zhao2,3 1 Texas A&M University, College Station, TX 2 Institute of Automation, Chinese Academy of Sciences, Beijing, China 3 University of Chinese Academy of Sciences, Beijing, China 4 Ant Group, Hangzhou, China chenpei@tamu.edu, taifeng.wang@antgroup.com {kliu, yubo.chen, jzhao}@nlpr.ia.ac.cn Abstract This paper proposes a new task regarding event reason extraction from document-level texts. Unlike the previous causality detection task, we do not assign target events in the text but only provide structural event descrip- tions, and such settings accord more with prac- tice scenarios. Moreover, we annotate a large dataset FinReason for evaluation, which pro- vides Reasons annotation for Financial events in company announcements. This task is chal- lenging because the cases of multiple-events, multiple-reasons, and implicit-reasons are in- cluded. In total, FinReason contains 8,794 documents, 12,861 financial events and 11,006 reason spans. We also provide the perfor- mance of existing canonical methods in event extraction and machine reading comprehen- sion on this task. The results show a 7 percent- age point F1 score gap between the best model and human performance, and existing methods are far from resolving this problem. 1 Introduction Why does the event happen? People are always eager to find the reasons for an event. Automati- cally extracting the causal explanations of the given events from texts is useful and important for com- mon users and downstream applications. For exam- ple, in the financial domain, returning the reasons of a concerned financial event in an Information Retrieval system can free analysts from reading the enormous company announcements and help investors make financial decisions. Previous work on event causality (Do et al., 2011; Riaz and Girju, 2013; Mirza and Tonelli, 2014; Caselli and Vossen, 2017) mainly focus on ∗Most of the work was done when the first author was a research engineer in the Institute of Automation, CAS. ID="PLEDGE_2168" ……2017年1月13日，小康控股将其持有的公司40,000,000股限售流通股质押给重庆农商 行…质押期限自2017年1月6日至质押登记解除日为止…公司股东张容于2016年12月13日 将其持有的公司5,353,400股限售流通股质押给中信建投证券股份有限公司，用于办理 股票质押式回购交易…小康控股本次股份质押是为自身融资提供股权质押担保…… ……On Jan 13th 2017, Kang Inc. declared its pledge of 40,000,000 shares to CRCB bank since Jan 1st 2016. Shareholder Rong Zhang pledged her holdings of 5,353,400 to CITICS Inc. from Dec. 13th 2016 for stock pledged repo transactions… Kang’s pledge aimed at providing guarantee for self-financing…… Reason 1 质押是为自身融资提供股权质押担保 aimed at providing guarantee for self-financing Reason 2 用于办理股票质押式回购交易业务 for stock pledged repo transactions NAME ORG NUM BEG Event 1 小康控股 Kang Inc. 重庆农商行 CRCB bank 40,000,000 2017-01-06 Event 2 张容 Rong Zhang 中信建投 CITICS Inc. 5,353,400 2016-12-13 + Figure 1: An example of reason extraction for struc- tural events from a document. We need to extract the textual spans from the document as reasons for the given structurally presented events. Here, we need to extract Reason 1 and Reason 2 for Event 1 and Event 2 respectively. the identification of causal relations …

Content truncated to 500 words