Abstract

entry into a NIST MT evaluation. Our entry to the 2008 evaluation mainly focused on establishing a competent baseline with a phrase-based system similar to (Och and Ney, 2004; Koehn et al., 2007). In a three-week effort prior to the evaluation, our attention focused on scaling up our system to exploit nearly all Chinese-English parallel data permissible under the constrained track, incorporating competitive language models into the decoder using Gigaword and Google n-grams, evaluating Chinese word segmentation models, and incorporating a document classifier as a pre-processing stage to the decoder. 4.1 Core MT engine algorithmic approach Our translation system contains four main components: a phrase-based translation system, a Chinese word segmenter, a text categorizer, and a truecaser. The purpose of the text categorizer is to classify each test document as either newswire or web, which allows us to run our phrase-based decoder with a set of log-linear parameters tuned on the predicted genre. The three latter components are presented in Section 4.2. 4.1.1 Parallel data Our translation models were trained using almost all of the constrained track Chinese-English data, leaving aside only the Multiple Translation Chinese corpus (LDC2006T04), OntoNotes (LDC2007T21), test data from previous NIST MT evaluations

Citations