Efficient Entity Candidate Generation for Low-Resource Languages

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 6429–6438 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 6429 Efficient Entity Candidate Generation for Low-Resource Languages Alberto Garc´ıa-Dur´an, Akhil Arora, Robert West EPFL, Switzerland agaduran@gmail.com {akhil.arora, robert.west}@epfl.ch Abstract Candidate generation is a crucial module in entity linking. It also plays a key role in multiple NLP tasks that have been proven to beneficially leverage knowledge bases. Nevertheless, it has often been overlooked in the monolingual English entity linking literature, as na¨ıve approaches obtain very good performance. Unfortunately, the existing approaches for English cannot be successfully transferred to poorly resourced languages. This paper constitutes an in-depth analysis of the candidate generation problem in the context of cross-lingual entity linking with a focus on low-resource languages. Among other contributions, we point out limitations in the evaluation conducted in previous works. We introduce a characterization of queries into types based on their difficulty, which improves the interpretability of the performance of different methods. We also propose a light-weight and simple solution based on the construction of indexes whose design is motivated by more complex transfer learning based neural approaches. A thorough empirical analysis on 9 real-world datasets under 2 evaluation settings shows that our simple solution outperforms the state-of-the-art approach in terms of both quality and efficiency for almost all datasets and query types. Keywords: Candidate generation, low-resource languages, evaluation pitfalls, efficiency, light-weight models 1. Introduction Candidate generation (Shen et al., 2014; Zhou et al., 2020) is the task of retrieving a short list of plausible candidate entities from a knowledge base (KB) for a given mention. This module is usually a fundamental part of a more elaborate pipeline as that of entity link- ing (Guo et al., 2013), coreference resolution (Zhang et al., 2019) or question answering (Bao et al., 2016). With few exceptions (Gillick et al., 2019), it is present in almost all existing entity linking techniques, where the subsequent entity disambiguation technique only has to evaluate a small set of entities from the KB, instead of all of them. Therefore, in addition to ob- taining a high recall—fraction of mentions where the true entity is present in the set of candidates—a de- sired characteristic of a candidate generator is to consti- tute a low computational overhead of the full pipeline. Cross-lingual entity linking (Tsai and Roth, 2016; Sil et al., 2018) associates mentions in non-English doc- uments to entities from a KB—typically the English Wikipedia. This poses several challenges that can of- ten be attributed to the amount of resources of the tar- get language, especially if it has low resources. Can- didate generation is deemed trivial in the English en- tity linking literature, as there exist resources (i.e. in- dexes/lookup tables (Spitkovsky and Chang, 2012)) ca- pable of providing high candidate recall. However, this is not the case in the cross-lingual setting. This is specially true when dealing with low-resource lan- guages (e.g., karelian or turkmen), wherein, moreover, the number …

Content truncated to 500 words