WikiCoref: An English Coreference-annotated Corpus of Wikipedia Articles

WikiCoref: An English Coreference-annotated Corpus of Wikipedia Articles Abbas Ghaddar, Philippe Langlais RALI-DIRO Montreal, Canada abbas.ghaddar@umontreal.ca, felipe@iro.umontreal.ca Abstract This paper presents WikiCoref, an English corpus annotated for anaphoric relations, where all documents are from the English version of Wikipedia. Our annotation scheme follows the one of OntoNotes with a few disparities. We annotated each markable with coreference type, mention type and the equivalent Freebase topic. Since most similar annotation efforts concentrate on very specific types of written text, mainly newswire, there is a lack of resources for otherwise over-used Wikipedia texts. The corpus described in this paper addresses this issue. We present a freely available resource we initially devised for improving coreference resolution algorithms dedicated to Wikipedia texts. Our corpus has no restriction on the topics of the documents being annotated, and documents of various sizes have been considered for annotation. Keywords: Annotated Corpus, Coreference Resolution, Wikipedia 1. Introduction In the last decade, coreference resolution has received an increasing interest from the NLP community, and became a standalone task in conferences and competitions due its role in applications such as Question Answering (QA), Infor- mation Extraction (IE), etc. This can be observed through, either the growth of coreference resolution systems vary- ing from machine learning approaches e.g (Haghighi and Klein, 2009) to rule based systems e.g. (Lee et al., 2013), or the large-scale of annotated corpora comprising different text genres and languages. Wikipedia1 is a very large multilingual, domain- independent encyclopedic repository. The English version, as of July 2015, contains more than 4M articles, thus providing a large coverage of knowledge resources. Wikipedia articles are highly structured and follow strict guidelines and policies. Not only are articles formatted into sections and paragraphs, moreover volunteer contrib- utors are expected to follow a number of rules2 (specific grammars, vocabulary choice and other language specifi- cations) that makes Wikipedia articles a text genre of its own. Over the past few years, Wikipedia imposed itself on coref- erence resolution systems as a semantic knowledge source, owing to its highly structured organization and especially to a number of useful reference features such as redirects, out links, disambiguation pages, and categories. Although the boost in English annotated corpora tagged with anaphoric coreference relations and attributes, none of them involve Wikipedia articles as its main component. This matter of fact motivated us to annotate Wikipedia documents for coreference, with the hope that it will fos- ter research dedicated to this type of text. We introduce WikiCoref, an English corpus, constructed purely from Wikipedia articles, with the main objective to balance top- 1https://www.wikipedia.org/ 2https://en.wikipedia.org/wiki/Wikipedia:Manual of Style ics and text size. This corpus has been annotated neatly by embedding state-of-the art tools (a coreference resolu- tion system as well as a Wikipedia/FreeBase entity detec- tor) that were used to assist manual annotation. This phase was then followed by a correction step to ensure fine qual- ity. Our annotation scheme is mostly similar to the one fol- lowed within the OntoNotes project (Pradhan et al., 2007), yet with some minor …

Content truncated to 500 words