The Parsing Algorithm of Translation Corresponding Tree (TCT) Grammar

In machine translation (MT), parsing acts as a kernel step to analyze and acquire the syntactic information of an input sentence for the purpose to reproduce the corresponding translation in target language according to the syntactic relationships between the source and target sentences. The parsing process is guided by a set of language formalism, and the design of such algorithm is highly depending on how the language information it represents, especially in the development of example-based machine translation (EBMT) system, where the foundmental language information are the set of translation examples. In this paper, a parsing algorithm is designed to parse the Translation Corresponding Tree (TCT) structure based on the augmented GLR algorithm. The TCT, as the examples representation schema, has been used to the annotation of bilingual text in EBMT system. In order to achieve in parsing the tree language based on GLR parsing algorithm, a TCT equivalent synchronous formalism based on the notation of context free grammar (CFG) is proposed. Where the feature properties of a TCT structure can be fully expressed in term of the proposed formalism, and which, as a result, can be parsed by any CFG parsers. In this paper, GLR algorithm is extended and adapted to this parsing task in the development of an EBMT system for Portuguese to Chinese machine translation.