WEXEA: Wikipedia EXhaustive Entity Annotation

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 1951–1958 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 1951 WEXEA: Wikipedia EXhaustive Entity Annotation Michael Strobl, Amine Trabelsi, Osmar R. Za¨ıane University of Alberta, Edmonton, Canada {mstrobl, atrabels, zaiane}@ualberta.ca Abstract Building predictive models for information extraction from text, such as named entity recognition or the extraction of semantic relationships between named entities in text, requires a large corpus of annotated text. Wikipedia is often used as a corpus for these tasks where the annotation is a named entity linked by a hyperlink to its article. However, editors on Wikipedia are only expected to link these mentions in order to help the reader to understand the content, but are discouraged from adding links that do not add any benefit for understanding an article. Therefore, many mentions of popular entities (such as countries or popular events in history), or previously linked articles, as well as the article’s entity itself, are not linked. In this paper, we discuss WEXEA, a Wikipedia EXhaustive Entity Annotation system, to create a text corpus based on Wikipedia with exhaustive annotations of entity mentions, i.e. linking all mentions of entities to their corresponding articles. This results in a huge potential for additional annotations that can be used for downstream NLP tasks, such as Relation Extraction. We show that our annotations are useful for creating distantly supervised datasets for this task. Furthermore, we publish all code necessary to derive a corpus from a raw Wikipedia dump, so that it can be reproduced by everyone. Keywords: Wikipedia, Knowledge Graphs, Relation Extraction, Information Extraction, Annotated Data, Distant Supervision 1. Introduction Understanding factual knowledge from text data is a cru- cial skill towards mastering several high-level NLP tasks, such as Response Generation for conversational agents (Lo- gan et al., 2019) or Question Answering (Yih et al., 2015). Elements of factual knowledge can include the following items: • What are the entities in context? • What is happening? Typically an event about how en- tities are interacting with or related to each other or what they are doing. (“John married Lisa“, “John was born in New York City”) • When is this happening? Events have a timestamp or a period of time is attached. (“John was born on Novem- ber 11th, 1976”) Apart from these three elements, text can also contain other numerical data, e.g. quantities with units, that express something of interest and are not just a sequence of distinct words as usually assumed in Language Modeling. In order to achieve the task of extracting factual knowledge from text or conversations, several subtasks have to be mas- tered (either pipelined or in an end-to-end fashion): 1. Entity Recognition (ER) (Lample et al., 2016): Typi- cally formulated as Named Entity Recognition (NER) task aiming to find all named entities in text. Different tagging schemes are possible, although the most com- mon scheme is the CoNLL-2003 scheme (Sang and De Meulder, 2003) with 4 classes: …

Content truncated to 500 words