Research On Obtaining English And Chinese Translation Examples On Shallow Parsing

Translation examples are the source of machine translation and the acquisition of it is the key technology for improving the quality of translation results. Translation examples are indispensable knowledge for further development in the field of machine translation. However, research on translation examples acquisition seems still weak. Working on acquisiting examples from parallel corpus is of great theoretical significance and practical potentiality.Current methods of acquisiting translation example automatically have many disadvantages. Lately, many researchers have been focusing on studing shallow parsing. These studies help to divide sentences into chunks. In this paper, chunking methods of Chinese and English language are used to preprocess the parallel corpus. On the basis of it, better translation examples can be acquired. This paper researches on obtaining translation examples from the following aspects.1.Basing on the knowledge of statistics and linguistics, this paper does researches on the rules of markers. Experiments show that the new proposed markers are much better than the former research on Chinese markers and also prove the feasibility of markers. Once having obtained the set of marker, we can part sentence for example just relying on the word tag. This research has its magnificent importance.2.This paper finds ways of aligning bilingual chunks for the results of chunking. From this aspect, the paper proposes four methods for the chunk alignment. We take on experiments and research either from English chunks or from Chinese chunks.3.On the basis of bilingual chunk results, this paper put up three aligning ways for them. We explore the difference and feasibility of the way completely relying on word alignment, the way completely relying on word dictionary, and the way of making use of both word alignment and dictionary.4.According to the comparison and analysis of the mentioned ways of obtaining translation examples, this paper proposes an algorithm of multi- subtracting and choosing for improving the quality of the examples. Experiments show that this way is a good one for obtaining translation examples. It also helps to introduce a new way of continual exploring research on this field.