Machine Translation: A Cognitive Linguistics Approach

This paper describes a novel approach to the construction of an interlinguabased MT system. This approach emerges from the developing field of Cognitive Linguistics. The objective of this paper is to address one of the main problems in MT systems (besides the source text ambiguity problem): the apparent complexity and irregularity of translation data. The author suggests that the key to a coherent account for some of the over-diversified translation data lies in the integration of an additional knowledge-base into the translation system: the knowledge of language usage. This knowledge is not part of the structure of language itself but derives from cognitive mechanisms which language acts upon. Cognitive Linguistics introduces a framework which gives a central role, in any understanding of semantics, to various kinds of schemas which are triggered by the language but which are not an explicit part of language itself. The paper examines some intriguing examples of English grammatical structures and their diversified translation into Hebrew. The analysis shows that by extracting the right schema from the source text, we can generate a better representation of the text's meaning which leads to easier and more accurate generation of the target text.