4.4 Effectiveness of Parallel Corpus
methods use pooling strategy to estimate the transla-
Preprocessing
tion probabilities. There are some clear trends in the
Question-answer pairs collected from Yahoo! an-
result of Table 4:
swers are very noisy, it is possible for translation
(1) Word-based translation language model
models to contain “unnecessary” translations. In this
(TransLM) significantly outperforms word-based
paper, we attempt to identify and decrease the pro-
translation model of Jeon et al. (2005) (row 1 vs. row
portion of unnecessary translations in a translation
Bạn đang xem 4. - TÀI LIỆU BÁO CÁO KHOA HỌC PHRASE BASED TRANSLATION MODEL FOR QUESTION RETRIEVAL IN COMMUNITY QUESTION ANSWER ARCHIVES PPT
