Open Information Extraction with Entity Focused Constraints

Findings of the Association for Computational Linguistics: EACL 2023, pages 1285–1296 May 2-6, 2023 ©2023 Association for Computational Linguistics Open Information Extraction with Entity Focused Constraints Prajna Upadhyay BITS Pilani Hyderabad Campus, Secunderabad, India prajna.u@hyderabad.bits-pilani.ac.in Oana Balalau and Ioana Manolescu Inria and Institut Polytechnique de Paris Palaiseau, France first.last@inria.fr Abstract Open Information Extraction (OIE) is the task of extracting tuples of the form (subject, predicate, object), without any knowledge of the type and lexical form of the predicate, the subject, or the object. In this work, we focus on improving OIE quality by exploiting domain knowledge about the subject and object. More precisely, knowing that the subjects and objects in sentences are often named entities, we explore how to inject constraints in the extraction through constrained inference and constraint-aware training. Our work lever- ages the state-of-the-art OpenIE6 platform, which we adapt to our setting. Through a carefully constructed training dataset and constrained training, we obtain a 29.17% F1-score improvement in the CaRB metric and a 24.37% F1-score improvement in the WIRe57 metric. Our technique has important applications – one of them is investigative journalism, where automatically extracting conflict-of-interest between scientists and fund- ing organizations helps understand the type of relations companies engage with the scientists. Our code and data are available at https: //github.com/prajnaupadhyay/ openie-with-entities 1 Introduction Open Information Extraction (OIE) is the task of extracting triples from unstructured corpora in a domain-independent manner. A triple consists of a subject, a relation, and an object. OIE has impor- tant applications, such as question answering (Lu et al., 2019), or automatically creating or extending knowledge bases (Bhutani et al., 2019). OIE is a challenging task, with the performance of state-of- the-art models varying from 88.5% F1 score (Wang The work was done when the first author was at Inria and Institut Polytechnique de Paris. et al., 2021) to 34% (Gashteovski et al., 2021), de- pending on the difficulty of the benchmark. When the named entities in a domain are known to be the subject/object of extractions, OIE should also identify relations between these entities. An important use case is automatically creating a knowledge base of relations between scientists and companies, i.e. identifying conflict-of-interest between the scientists and funding bodies, where the named entities are the names of scientists and companies, and the relation describes the conflict of interest between them. Clustering these relation phrases, such as received a research gift from, received speaker fees or consults for helps analyze the relationships that companies engage with the scientists. These relations are crucial to understanding scientists’ positions on health issues (Oreskes and Conway, 2010) in investigative journalism. However state-of-the-art OIE models do not always retain named entities in the extractions, for example, given the sentence “Shahrad Taheri received funding for research through a grant from Cambridge Weight Plan”, an OIE tool (Kolluru et al., 2020a) returns ⟨Shahrad Taheri, received, funding for research⟩. While this extraction correctly identifies the subject of the triple, the quality of the predicate and object could be improved as …

Content truncated to 500 words