KGPool: Dynamic Knowledge Graph Context Selection for Relation Extraction

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 535–548 August 1–6, 2021. ©2021 Association for Computational Linguistics 535 KGPool: Dynamic Knowledge Graph Context Selection for Relation Extraction Abhishek Nadgeri1,2, Anson Bastos1,3, Kuldeep Singh1,5, Isaiah Onando Mulang’1,7 Johannes Hoffart6, Saeedeh Shekarpour4, and Vijay Saraswat6 1Zerotha Research, 2RWTH Aachen, 3IIT Hyderabad, 4University of Dayton 5Cerence GmbH, 6Goldman Sachs, 7IBM Research, Africa abhishek.nadgeri@rwth-aachen.de, cs20resch11002@iith.ac.in kuldeep.singh1@cerence.com, mulang.onando@ibm.com sshekarpour1@udayton.edu,{johannes.hoffart,vijay.saraswat}@gs.com Abstract We present a novel method for relation extrac- tion (RE) from a single sentence, mapping the sentence and two given entities to a canon- ical fact in a knowledge graph (KG). Espe- cially in this presumed sentential RE setting, the context of a single sentence is often sparse. This paper introduces the KGPool method to address this sparsity, dynamically expanding the context with additional facts from the KG. It learns the representation of these facts (en- tity alias, entity descriptions, etc.) using neu- ral methods, supplementing the sentential con- text. Unlike existing methods that statically use all expanded facts, KGPool conditions this expansion on the sentence. We study the ef- ficacy of KGPool by evaluating it with dif- ferent neural models and KGs (Wikidata and NYT Freebase). Our experimental evaluation on standard datasets shows that by feeding the KGPool representation into a Graph Neural Network, the overall method is significantly more accurate than state-of-the-art methods. 1 Introduction Knowledge graphs (KGs) are the foundation for many downstream applications and are growing ever larger. However, due to the sheer volume of knowledge and the world’s dynamic nature where new entities emerge and unknown facts about them are learned, KGs need to be continuously updated. Distantly supervised Relation Extraction (RE) is an important KG completion task aiming at finding a semantic relationship between two en- tities annotated on the unstructured text with re- spect to an underlying knowledge graph (Ye and Ling, 2019). In the literature, researchers mainly studied two variants in the RE: 1) multi-instance RE and 2) sentential RE. The multi-instance RE assumes that in a given bag of sentences, if two entities participate in a relation, there exists at least one sentence with these two entities, which wd:Q266569 wd:Q568631 wdt:P106 "Animator" "occupation" "Marc Davis" " American artist and animator" xsd:string "Marc Fraser Davis" xsd:string "Fraser Davis" xsd:string Human xsd:string "cartoonist" xsd:string "person who makes animated sequences out of still images" filmmaking occupation She was married to Marc Davis, a Disney animator, and Imagineer KG Sentence Alias Instance of Description Figure 1: Illustration of Knowledge Graph Context as- sociated with the annotated entities in a sentence. Here, entity aliases do not play any role in the understanding of the sentence for finding the KG relation. a asentence taken from (Sorokin and Gurevych, 2017) may contain the target relation (Riedel et al., 2010; Vashishth et al., 2018). In this setting, researchers aim to incorporate contextual signals from the pre- vious occurrences of an entity pair into the neu- ral models to support relation extraction (Ye and Ling, 2019; Xu and Barbosa, 2019; Wu et al., …

Content truncated to 500 words