EFLLex: A Graded Lexical Resource for Learners of English as a Foreign Language

EFLLex: A Graded Lexical Resource for Learners of English as a Foreign Language D¨urlich, L.1,2, Franc¸ois, T.2,3 1 FAU Erlangen-N¨urnberg, 2 CENTAL/IL&C, University of Louvain 3 Charg´e de recherche FNRS luise.duerlich@fau.de, thomas.francois@uclouvain.be Abstract This paper introduces EFLLex, an innovative lexical resource that describes the use of 15,280 English words in pedagogical materials across the proficiency levels of the European Framework of Reference for Languages. The methodology adopted to produce the resource implies the selection of an efficient part-of-speech tagger, the use of a robust estimator for frequency computation and some manual post-editing work. The content of the resource is described and compared to other vocabulary lists (MRC and BNC) and to a reference pedagogical resource: the English Vocabulary Profile. Keywords: CEFR-graded lexicon, English as a foreign language, vocabulary 1. Introduction The strong relation between vocabulary knowledge and reading comprehension has been thoroughly researched and confirmed (Laufer, 1992). Second language acquisition (SLA) research has established that readers should know between 95% and 98% of the words in a text to adequately comprehend it (Laufer and Ravenhorst-Kalovski, 2010). Given that the larger the vocabulary, the better the com- prehension of texts, foreign language curriculums should stress the need to teach new vocabulary items. A larger vocabulary indeed means being able to understand a larger scope of texts. However, analyses of large corpora indicate that, from kindergarten through college, native speakers en- counter approximately 150,000 different words (Zeno et al., 1995). For learners, it is impossible to pick up all of them. Designers of foreign language curriculums, publishers of educational materials and textbooks, or even teachers are thus faced with the issue of identifying the most important words to teach at each stage of the learning process. The most common answers to that challenge have been (1) to use frequency lists obtained from a large corpus of texts intended for a learner audience and (2) to rely on expert knowledge, such as teacher expertise or linguists’ recommendations. Following the first approach, frequency lists have been derived from various well-known corpora of English. The first significant vocabulary list for En- glish included 20,000 words and was laboriously produced by Thorndike (1921) without the help of any computer. It was later extended to 30,000 words by Thorndike and Lorge (1944). The first computational list was obtained by Kuˇcera and Francis (1967) from the Brown corpus and has a large influence in education and psychology. More recently, other lists have been developed from larger cor- pora, such as the CELEX database (Baayen et al., 1993), the Zeno list (Zeno et al., 1995), the list based on the British National Corpus (BNC) (Leech et al., 2001), or SUBLEX (Brysbaert and New, 2009). The main shortcomings of such lists for L2 education are that (1) they represent the native distribution of words, which is not fully compatible with the distribution of words in books and textbooks intended for L2 learners; (2) they do not specify at which proficiency level a given word is supposed to be learned. The …

Content truncated to 500 words