A Multilingual Wikified Data Set of Educational Material

A Multilingual Wikified Data Set of Educational Material Iris Hendrickx1, Eirini Takoulidou3, Thanasis Naskos3, Katia Lida Kermanidis3, Vilelmini Sosoni2, Hugo de Vos1, Maria Stasimioti2, Menno van Zaanen4, Panayota Georgakopoulou6, Markus Egg5, Valia Kordoni5, Maja Popovic5, Antal van den Bosch1 1Centre for Language Studies, Radboud University,Nijmegen, Netherlands 1Department of Foreign Languages, Translation and 2Interpreting, Ionian University, Corfu, Greece 3Department of Informatics, Ionian University, Corfu, Greece 4Department of Communication and Information Sciences, Tilburg University, The Netherlands 5Department of English Studies, Humboldt−Universit¨at zu Berlin, Germany 6Deluxe Entertainment Services Group, Athens, Greece i.hendrickx@let.ru.nl;rinoulit@gmail.com;anaskos@ionio.gr; vilelmini@hotmail.com; kerman@ionio.gr; stasimioti.maria@gmail.com; mvzaanen@uvt.nl; yota.georgakopoulou@bydeluxe.com; kordonie@anglistik.hu-berlin.de; maja.popovic@anglistik.hu-berlin.de; markus.egg@anglistik.hu-berlin.de;a.vandenbosch@let.ru.nl Abstract We present a parallel wikified data set of parallel texts in eleven language pairs from the educational domain. English sentences are lined up to sentences in eleven other languages (BG, CS, DE, EL, HR, IT, NL, PL, PT, RU, ZH) where names and noun phrases (entities) are manually annotated and linked to their respective Wikipedia pages. For every linked entity in English, the corresponding term or phrase in the target language is also marked and linked to its Wikipedia page in that language. The annotation process was performed via crowdsourcing. In this paper we present the task, annotation process, the encountered difficulties with crowdsourcing for complex annotation, and the data set in more detail. We demonstrate the usage of the data set for Wikification evaluation. This data set is valuable as it constitutes a rich resource consisting of annotated data of English text linked to translations in eleven languages including several languages such as Bulgarian and Greek for which not many LT resources are available. Keywords: MOOCs, crowdsourcing, wikification 1. Introduction TraMOOC (Translation for Massive Open Online Courses) is an EU-funded project that aims to improve access to ed- ucational material in MOOCs by providing dedicated ma- chine translation (MT) solutions. The educational con- tent, including video lecture subtitles, forum posts, and quiz questions and answers, is translated from English into eleven European and BRIC languages: Bulgarian, Chinese, Croatian, Czech, Dutch, German, Greek, Italian, Polish, Portuguese, and Russian. Here, we focus on the collection of parallel texts in eleven language pairs that are annotated and linked with respect to the entities occurring in the text. We consider names and noun phrases that contain the topical information about a text as entities. To identify such entities, information from Wikipedia is used. Each entity in the text is linked to its corresponding Wikipedia page in the respective language. The fine-grained information encapsulated in this data set can be used for a wide range of applications. We use the data set for an in-depth analysis and evaluation of machine translation output. In addition to word-based evaluations (e.g., BLEU), a semantic evaluation can be performed, as the links to the Wikipedia pages in the various target lan- guages provide an additional source of information. Such implicit MT evaluation aims to judge the MT quality be- tween source and target language without using an explicit (manual) translation step. Alternatively, the data set is suited for other multilingual …

Content truncated to 500 words