Abstract:
This thesis proposes a technique to rearrange words in an English sentence to achieve the correct semantic translation into the corresponding Thai sentence structure. Each phrase in an English sentence is coded by substituting its linguistic features such as case grammars, grammatical functions, types of English construction and verb features in forms of binary numbers. Similarly, a Thai sentence structure is obtained from the grammatical permutation of the English phrases together with Thai decorating words in forms of binary numbers. Hence the coded phrase can be viewed as a binary vector. Based on this coding scheme, this problem is, therefore, transformed to the problem of seeking a good function for mapping a binary vector representing an English sentence to abinary vector representing a Thai sentence structure. Due to the flexibility to approximate any function, we propose to deploy a supervised feedforward neural network to represent this mapping function.