Unsupervised Entity Linking with Guided Summarization and Multiple-Choice Selection

Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 9394 - 9401 December 7-11, 2022 ©2022 Association for Computational Linguistics Unsupervised Entity Linking with Guided Summarization and Multiple-Choice Selection Young-Min Cho Li Zhang Chris Callison-Burch University of Pennsylvania {jch0, zharry, ccb}@seas.upenn.edu Abstract Entity linking, the task of linking potentially ambiguous mentions in texts to correspond- ing knowledge-base entities, is an important component for language understanding. We address two challenge in entity linking: how to leverage wider contexts surrounding a men- tion, and how to deal with limited training data. We propose a fully unsupervised model called SumMC that first generates a guided summary of the contexts conditioning on the mention, and then casts the task to a multiple-choice problem where the model chooses an entity from a list of candidates. In addition to evalu- ating our model on existing datasets that focus on named entities, we create a new dataset that links noun phrases from WikiHow to Wikidata. We show that our SumMC model achieves state- of-the-art unsupervised performance on our new dataset and on existing datasets. 1 Introduction Entity linking (EL) is an important Natural Lan- guage Processing (NLP) task that associates am- biguous mentions to corresponding entities in a knowledge base (KB, also called knowledge graph). EL is a crucial component of many NLP applica- tions, such as question answering (Yih et al., 2015) and information extraction (Hoffart et al., 2011). Although there have been significant and contin- uous developments of EL, most work requires suffi- cient labeled data and a well-developed KB (Zhang et al., 2021; Mulang’ et al., 2020; van Hulst et al., 2020; Raiman and Raiman, 2018). However, many real-world applications, especially those in specific domains, suffer from scarcity of both training data and a fully-populated KB. Previous research has tackled this problem by learning EL models with- out data labeled entity links, but requires indirect supervision in the form of textual descriptions at- tached to entities in KBs, drawn from sources such as Wikipedia (Cao et al., 2017; Logeswaran et al., 2019). However, such descriptions may not be Figure 1: Example of an Entity Linking problem. available in KBs in low-resource domains such as medicine or law. Thus, we focus on fully unsu- pervised EL, which only has access to the entities’ names and their KB relations like subclass-of (Le and Titov, 2019; Arora et al., 2021). One challenge of unsupervised EL is leveraging useful information from potentially noisy and mis- leading context (Pan et al., 2015). Specifically, a local context (the sentence containing the mention) may not be sufficient for disambiguating the target mention without the global context (other sentences in the document). For example, in Figure 1, the tar- get mention ‘band’ cannot be disambiguated solely with the local context “This band is so lovely”, but needs to consider the global context that also in- cludes “I can’t wait for my wedding.” To address this problem, we introduce an unsu- pervised approach to EL that builds on the …

Content truncated to 500 words