Fun with Filtering French

Fun with Filtering French Alexandra L. Uitdenbogerd RMIT University Melbourne, Victoria, Australia Abstract Early use of corpora for language learn- ing has included analysis of word usage via concordancing. In addition, some at- tempts have been made to use readabil- ity criteria for recommending reading to learners. In this paper we discuss var- ious tools and approaches for enhanced language learning support, including dif- ferent methods of filtering text based on vocabulary and grammatical criteria. We demonstrate the effects of various criteria on the retrieval of text, assuming the user is English-speaking and learning French. Filtering text based on a small vocabu- lary of frequently occurring words, a set of English-French cognates and named en- tities, and high coverage criteria, results in the retrieval of short readable extracts from French literature. We expect that text available from the web may yield many more documents of appropriate readabil- ity. 1 Introduction There is a considerable need for people to learn and become proficient in foreign languages: the majority of scientific discourse is published in English; students travel to different countries to study; people migrate for career opportunities. Language skills are often divided into four com- munication tasks: listening, speaking, reading, and writing. Each of these skills can be developed and practised separately to a certain extent. Im- proving reading skill in a language would clearly involve devoting a substantial amount of time to reading. It has been demonstrated that extensive read- ing at a comfortable level in a foreign language is more effective for improving language acquisi- tion than intensive reading at more difficult lev- els (Bell, 2001). Therefore there is a need for reading material at multiple language skill levels to allow learners to practise. Some publishers pro- vide graded reading books targeted at the foreign language learner, with levels indicated either by an assumed base vocabulary size, a standard level such as that defined by the Common European Framework of Reference for Languages (COE, 2003), year of study, or an unexplained level struc- ture. The simplest graded readers based on vocab- ulary size that we have seen use a base vocabulary of 150 words. Beginner readers tend to be much shorter in length than those for advanced learn- ers. For example, the level 1 readers in the Bib- liobus Collection A are approximately 150 words in length, in a comic book format, consisting of two short stories (Cowling, 1982). Several researchers independently proposed the idea of retrieving reading material from the Web based on its readability for the purpose of reading practice or study (Collins-Thompson and Callan, 2004a; Ghadirian, 2002; Katz and Bauer, 2001; Nilsson and Borin, 2002; Uitdenbogerd, 2003). This motivated some new studies of measur- ing readability (Collins-Thompson and Callan, 2004b; Schwarm and Ostendorf, 2005; Si and Callan, 2001) that are more sophisticated than was possible in the initial period of readability research (Bormuth, 1966; Chall and Dale, 1995; Cornaire, 1988; Granowskey and Botel, 1974; Klare, 1974). Our earlier work demonstrated that simple techniques are still very …

Content truncated to 500 words