A machine learning-based approach to missing preposition detection

抄録

The number of learners of English as a foreign language is estimated to be more than a billion, and is increasing. For this reason, various tools that help them detect and correct writing errors have been developed in the field of natural language processing. For example, spelling error correction tools have achieved high accuracy and have been widely used in the world. These tools are used not only for English learners, but also for other natural language processing systems, for instance, to improve outputs of machine translation systems. However, many other tasks, including grammar checking, are not currently done very effectively, and their development is in progress. n this paper, we propose a missing preposition detection system which exploits syntactic information provided by an English parser. The information enables us to focus on a location possibly lacking a preposition, and can also be utilized as a machine learning feature for determining whether the location really requires a preposition or not. In the experiments, we examine the effectiveness of our systems on English learner corpus, Konan-JIEM Learner Corpus, by comparing the detection accuracy with previous research.