Aligning Wikipedia with WordNet:a Review and Evaluation of Different Techniques

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 4851–4858 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 4851 Aligning Wikipedia with WordNet: a Review and Evaluation of Different Techniques Antoni Oliver Universitat Oberta de Catalunya (UOC) Barcelona (Catalonia-Spain) aoliverg@uoc.edu Abstract In this paper we explore techniques for aligning Wikipedia articles with WordNet synsets, their successful alignment being our main goal. We evaluate techniques that use the definitions and sense relations in Wordnet and the text and categories in Wikipedia articles. The results we present are based on two evaluation strategies: one uses a new gold and silver standard (for which the creation process is explained); the other creates wordnets in other languages and then compares them with existing wordnets for those languages found in the Open Multilingual Wordnet project. A reliable alignment between WordNet and Wikipedia is a very valuable resource for the creation of new wordnets in other languages and for the development of existing wordnets. The evaluation of alignments between WordNet and lexical resources is a difficult and time-consuming task, but the evaluation strategy using the Open Multilingual Wordnet can be used as an automated evaluation measure to assess the quality of alignments between these two resources. Keywords: WordNet, Wikipedia, sense alignment 1. Introduction In this paper we address the task of creating alignments between Princeton WordNet (PWN 3.0) synsets and En- glish Wikipedia articles. In principle the alignment of monosemic words should be relatively easy, as a simple comparison of the word forms would be sufficient. The problem arises with the granularity of senses involved when aligning polysemic words, i.e. words attached to several senses. One given word may be considered monosemic in one of the resources and polysemic in the other resource; or a given word may even be considered monosemic in both resources but refer to different senses in each resource. The task of aligning words that are polysemic in both resources is even more difficult. Wikipedia1 is a collaborative multilingual encyclopaedia that uses the wiki format. All Wikipedia’s contents can be downloaded as XML dumps containing the code in wiki format. Extracting the content and the associated informa- tion from the Wikipedia articles is a difficult task, but sev- eral parsers are available. For example, WikiExtractor 2, a Python script that can extract the text and other informa- tion from articles and give an XML dump; or JWPL3 (Java Wikipedia Library), an open-source Java-based API that gives access to all the information contained in Wikipedia. From the Mediawiki download page we can also download SQL table dumps containing information from Wikipedia articles such as their titles, categories and interlanguage links. DBpedia4 (Lehmann et al., 2015) is a project aiming to extract structured content from the Wikipedia project. This structured information can be downloaded in several database file formats. In our experiments we used DB- pedia and the following tables were processed: labels, 1www.wikipedia.org 2https://github.com/attardi/wikiextractor 3https://www.ukp.tu-darmstadt.de/software/jwpl 4http://wiki.dbpedia.org/ short abstracts, long abstracts, article categories and inter- …

Content truncated to 500 words