A French Corpus and Annotation Schema for Named Entity Recognition and Relation Extraction of Financial News

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2293–2299 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2293 A French Corpus and Annotation Schema for Named Entity Recognition and Relation Extraction of Financial News Ali Jabbari, Olivier Sauvage, Hamada Zeine, Hamza Chergui Skaizen Group Paris, France {ajabbari, osauvage, hzeine, hchergui}@skaizengroup.fr Abstract In financial services industry, compliance involves a series of practices and controls in order to meet key regulatory standards which aim to reduce financial risk and crime, e.g. money laundering and financing of terrorism. Faced with the growing risks, it is imperative for financial institutions to seek automated information extraction techniques for monitoring financial activities of their customers. This work describes an ontology of compliance-related concepts and relationships along with a corpus annotated according to it. The presented corpus consists of financial news articles in French and allows for training and evaluating domain-specific named entity recognition and relation extraction algorithms. We present some of our experimental results on named entity recognition and relation extraction using our annotated corpus. We aim to furthermore use the the proposed ontology towards construction of a knowledge base of financial relations. Keywords: Annotation Schema, Named Entity Recognition, Relation Extraction 1. Introduction Strict regulatory regimes mandate financial institutions (e.g. banks, insurance companies, investment firms) to rig- orously monitor their customers’ financial activities. Oper- ating in fear of massive fines in case of undetected viola- tion of the regulations, these institutes have set up a proce- dure called Know Your Customer (KYC). Today KYC pro- cedure is a time consuming effort of collecting information through declarative questionnaires and watchlist lookups. Furthermore, each customer’s profile need periodical re- views to ensure the accuracy of the provided information. The regulations in force also evolve over time and legisla- tors do not specify a standard of sufficient information in order to prevent financial institutes from enforcing a bare minimum control. This has led financial institutions to seek as much data as possible on their clients’ financial activi- ties. In this context, automated solutions can help financial institutions to continuously update the information that can impact the assessment of risk profiles, e.g. ownership and managerial changes and company’s domains of activity. This goal can be achieved by means of a knowledge base (KB) of the financial relations of the clients which benefits from population through automated information extraction. For example, the news headline “Jean-Dominique Senard appointed as Renault Chairman" announces a financial event (managerial change) and thus, implies that the exist- ing KYC records of those entities need to be updated. We would therefore like to extract the entities Jean-Dominique Senard as a Person, Renault as Company, and chairman as a Role (Named Entity Recognition). We would also like to extract the relation between the person and his role, and that between the role and the affiliated company (Relation Ex- traction). Extracting such information from unstructured data raises several challenges. Among other determining factors, the performance of named entity recognition …

Content truncated to 500 words