Leveraging Native Data to Correct Preposition Errors in Learners’ Dutch

Leveraging Native Data to Correct Preposition Errors in Learners’ Dutch Lennart Kloppenburg and Malvina Nissim Center for Language and Cognition Groningen – Language Technology Rijksuniversiteit Groningen, The Netherlands l.a.m.kloppenburg@student.rug.nl, m.nissim@rug.nl Abstract We address the task of automatically correcting preposition errors in learners’ Dutch by modelling preposition usage in native language. Specifically, we build two models exploiting a large corpus of Dutch. The first is a binary model for detecting whether a preposition should be used at all in a given position or not. The second is a multiclass model for selecting the appropriate preposition in case one should be used. The models are tested on native as well as learners data. For the latter we exploit a crowdsourcing strategy to elicit native judgements. On native test data the models perform very well, showing that we can model preposition usage appropriately. However, the evaluation on learners’ data shows that while detecting that a given preposition is wrong is doable reasonably well, detecting the absence of a preposition is a lot more difficult. Observing such results and the data we deal with, we envisage various ways of improving performance, and report them in the final section of this article. Keywords: Preposition correction, language modelling, learners’ data, Dutch 1. Introduction Computer-assisted language learning is a field where lan- guage technology is combined with human language ac- quisition. To make it possible, we need systems that are able to recognise errors and suggest corrections. However, in spite of recent campaigns on automatic error correction (Ng et al., 2013; Ng et al., 2014), and in spite of some no- table exceptions (Nicholls, 2003) the lack of large amounts of error-annotated data remains a bottleneck in the develop- ment and evaluation of systems that support language learn- ing, especially for L2s other than English. One obvious solution is creating error-aware resources to build and test robust models of learners’ language use (Han et al., 2010), which has however substantial costs, especially in terms of human effort. Another way of by- passing this shortage of data is to exploit native data, which can be considered as having gold labels of correct usage of language (Han et al., 2006; Tetreault and Chodorow, 2008; Gamon et al., 2008; De Felice and Pulman, 2009). This latter strategy is what we adopt in this work, focusing on preposition errors in learners of Dutch. As far as we are aware, this is the first such work for this language. It has been noted that in many languages prepositional constructions are the most difficult for foreign language learners, because of their versatile and often ambiguous character (Dale et al., 2012). Although we have no spe- cific figures on this kind of error in learners of Dutch, we know that prepositions are used extremely frequently and that large corpora of native Dutch are available, which can be exploited to build models of correct preposition usage. In this paper we describe a two-stage approach to the detection and correction of preposition errors in essays written …

Content truncated to 500 words