Syntax-Directed Machine Translation of Natural Language: Effect of Garden Path Phenomenon on Sentence Structure

The potential of current machine translation (MT) of natural languages is discussed by comparing the output of embedded structure sentence, ambiguous sentence and garden path sentence. The syntax-oriented MT is a kind of literal translation which may be regarded as an idealized competence of today's practical MT. The above comparison shows that the main reason for the incompleteness of current practical MT systems is the insufficient ability to treat structural ambiguousness of sentences. With the help of Recursive Transition Network (RTN), N-S Flowchart techniques for structured programming and Context-Free Grammar (CFG), the translation examples present that, without understanding the total meaning of the source sentence, it is quite difficult to manipulate the garden path phenomenon in sentence structure. Garden path effect on sentence structure usually originates from syntactic differences, so that the syntax-free computational translation of these sentences is a very difficult problem. However, encouraging evidence is provided that the principal technologies of today's hybrid-knowledge-based MT have sufficient potential for producing barely acceptable translation. Garden path sentence possesses the features of processing breakdown and backtracking, which bring lingering effect on automatic processing. Only on the basis of hybrid knowledge databases of MT system can the suffering from garden path effect on sentence structure be alleviated and can robustness of MT system be enhanced.