A High Precision Pipeline for Financial Knowledge Graph Construction

Proceedings of the 28th International Conference on Computational Linguistics, pages 967–977 Barcelona, Spain (Online), December 8-13, 2020 967 A High Precision Pipeline for Financial Knowledge Graph Construction Sarah Elhammadi, Laks V.S. Lakshmanan, Raymond Ng, Michael Simpson University of British Columbia, Vancouver, Canada shammadi,laks,rng,mesimp@cs.ubc.ca Baoxing Huai, Zhefeng Wang, Lanjun Wang∗ Huawei Technologies Co. Ltd., Hangzhou, China huaibaoxing,wangzhefeng,lanjun.wang@huawei.com Abstract Motivated by applications such as question answering, fact checking, and data integration, there is significant interest in constructing knowledge graphs by extracting information from unstruc- tured information sources, particularly text documents. Knowledge graphs have emerged as a standard for structured knowledge representation, whereby entities and their inter-relations are represented and conveniently stored as (subject, predicate, object) triples in a graph that can be used to power various downstream applications. The proliferation of financial news sources reporting on companies, markets, currencies, and stocks presents an opportunity for extracting valuable knowledge about this crucial domain. In this paper, we focus on constructing a knowl- edge graph automatically by information extraction from a large corpus of financial news articles. For that purpose, we develop a high precision knowledge extraction pipeline tailored for the fi- nancial domain. This pipeline combines multiple information extraction techniques with a finan- cial dictionary that we built, all working together to produce over 342,000 compact extractions from over 288,000 financial news articles, with a precision of 78% at the top-100 extractions. The extracted triples are stored in a knowledge graph making them readily available for use in downstream applications. 1 Introduction Knowledge graphs (KG) have lately emerged as a de facto standard for knowledge representation in the Semantic Web, whereby knowledge is expressed as a collection of “facts”, represented in the form (subject, predicate, object) (SPO) triples, where subject and object are entities and predicate is a relation between those entities. This collection can be conveniently stored, queried, and maintained as a graph, with the entities modeled as vertices and relations as links or directed edges. Driven by applications such as question answering, fact checking, information search, data integration and recommender systems, there is tremendous interest in extracting high quality knowledge graphs by tapping various data sources (Ji et al., 2020; Noy et al., 2019; Dong et al., 2014; Shortliffe, 2012; Lehmann et al., 2015). Over the years, a number of cross domain KGs have been created including DBpedia (Lehmann et al., 2015), YAGO (Suchanek et al., 2007), Freebase (Bollacker et al., 2008) and NELL (Mitchell et al., 2018), covering millions of real world entities and thousands of relations, across different domains. Recently, there has been a growing interest in generating domain targeted structured representations of financial and business entities and how they are related to each other. Crunchbase 1 curated a knowledge base (KB) through partnerships with companies and data experts covering 100,000+ business entities including companies and investors, but covering only a few types of business transactions (i.e., relations) such as acquisitions and funding rounds. The work by (Benetka et al., 2017) attempted to address this limitation by developing a …

Content truncated to 500 words