Sentence-level evaluation using co-occurences of N-grams

Abstract

This work presents an evaluation method of Greek sentences with respect to word order errors. The evaluation method is based on words' reordering and choosing the version that maximizes the number of trigram hits according to a language model. The new parameter of the proposed technique concerns the incorporation of unigram probability. This probability corresponds to the frequency of each unigram to be posed in the first and in the last position of the training set sentences. The comparative advantage of this method is that it works with a large set of words, and avoids the laborious and costly process of collecting word order errors for creating error patterns. © Springer-Verlag Berlin Heidelberg 2008.