Overview of the 2015 ALTA Shared Task: Identifying French Cognates in English Text

Overview of the 2015 ALTA Shared Task: Identifying French Cognates in English Text Laurianne Sitbon Queensland University of Technology (QUT) l.sitbon@qut.edu.au Diego Molla Macquarie University, Sydney, Australia diego.molla- aliod@mq.edu.au Haoxing Wang Queensland University of Technology (QUT) haoxing.wang@hdr. qut.edu.au Abstract This paper presents an overview of the 6th ALTA shared task that ran in 2015. The task was to identify in English texts all the potential cognates from the per- spective of the French language. In other words, identify all the words in the Eng- lish text that would acceptably translate into a similar word in French. We present the motivations for the task, the descrip- tion of the data and the results of the 4 participating teams. We discuss the re- sults against a baseline and prior work. 1 Introduction Because many languages have evolved from a shared source language (e.g. Indo-European lan- guages), many words in their vocabularies are the same or are very similar. Additionally, global communications have facilitated the transfer of words from one language to another in modern languages. As a result, when learning a related language, a learner’s native language can support the acquisition and understanding of vocabulary words that are identical or similar in both lan- guages. A vocabulary word is a spelling associated to a particular meaning. Such pairs of identical or similar words that also share meaning across two languages are referred to as cognates. Definitions can vary in the level of similarity (exact or simi- lar spelling, exact or similar pronunciation, or both). So far, research on detecting cognates has focused on being able to identify pairs of cog- nates in lists of presented pairs of words. In contrast, in this shared task we use the no- tion of potential cognate in a target language with reference to a source language: a word in the target language that could be translated by a similar word in the source language such that these words form a cognates pair. Being able to identify these potential cognates in texts could provide technologies to extract easy to under- stand sentences and could support measures of reading difficulty (Uitdenbogerd, 2005) which can in turn be embedded in ranking information retrieval results or in sentence selection for summarization. In 2015, the sixth Australasian Language Technology Association (ALTA) shared task was set to identify in English texts all the poten- tial cognates from the perspective of the French language. A total of 6 teams registered to the competition, with 4 teams submitting their re- sults. In this paper we present some background for the task, describe the dataset and contrast the results of the participants against baselines and previous work. Section 2 presents some back- ground and prior work, Section 3 presents the task, the dataset and the evaluation measures. Section 4 provides the results of the participants. Section 5 discusses the results and future work. 2 Cognates identification and detection Cognates are pairs of words that are similar in spelling/pronunciation as well as meaning in two languages. …

Content truncated to 500 words