CLEEK: A Chinese Long-text Corpus for Entity Linking

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2026–2035 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2026 CLEEK: A Chinese Long-text Corpus for Entity Linking Weixin Zeng†, Xiang Zhao∗†‡, Jiuyang Tang†‡, Zhen Tan†, Xuqian Huang† †Science and Technology on Information Systems Engineering Laboratory, National University of Defense Technology, China ‡Collaborative Innovation Center of Geospatial Technology, China Abstract Entity linking, as one of the fundamental tasks in natural language processing, is crucial to knowledge fusion, knowledge base construction and update. Nevertheless, in contrast to the research on entity linking for English text, which undergoes continuous development, the Chinese counterpart is still in its infancy. One prominent issue lies in publicly available annotated datasets and evaluation benchmarks, which are lacking and deficient. In specific, existing Chinese corpora for entity linking were mainly constructed from noisy short texts, such as microblogs and news headings, where long texts were largely overlooked, which yet constitute a wider spectrum of real-life scenarios. To address the issue, in this work, we build CLEEK, a Chinese corpus of multi-domain long text for entity linking, in order to encourage advancement of entity linking in languages besides English. The corpus consists of 100 documents from diverse domains, and is publicly accessible. Moreover, we devise a measure to evaluate the difficulty of documents with respect to entity linking, which is then used to characterize the corpus. Additionally, the results of two baselines and seven state-of-the-art solutions on CLEEK are reported and compared. The empirical results validate the usefulness of CLEEK and the effectiveness of proposed difficulty measure. Keywords: Entity linking, Named entity disambiguation, Data annotation 1. Introduction With the exponential proliferation of unstructured data on the Internet, automatic extraction of valuable information from raw data becomes increasingly crucial. As one of the fundamental steps of bridging raw text and regularized knowledge, entity linking plays an indispensable role in various knowledge-centric tasks, such as knowledge fusion, knowledge base construction, etc. Entity linking (EL), or entity disambiguation, aims at map- ping ambiguous mentions in text to the true entities in a target knowledge base (KB). Entities are unique identifiers of objects, while mentions, as the surface forms of entities, usually possess various appearances. An instance of EL is depicted in Figure 1: there are three mentions underlined in the text, namely, The Bulls, Hinrich and Chandler. For each mention, EL first retrieves candidate entities from a target KB. As for mention The Bulls, the candidate enti- ties include the basketball team Chicago Bulls, the animal Bull, and possibly many others. Subsequently, EL selects the true entity out of the candidates. For example, the entity Chicago Bulls is the true entity for the mention The Bulls. EL is intrinsically non-trivial, since the diverse forms of mentions render it challenging to generate possible can- didate entities, let alone selecting true entities out of a group of similar candidates. Over the recent years, en- deavours have been devoted to designing accurate and ef- ficient EL …

Content truncated to 500 words