Machine translation using word connectivity relationships
Yasuhiro Suzuki, Koji Tochinai · Systems and Computers in Japan · 1997
The improvement in the performance of computers has prompted an increase in machine translation (MT) research. In general, MT methods are rule-based and convert the source language into the target language based on information obtained by morphologic, syntactic, and semantic analysis. As an alternative approach, we have developed an MT method taking advantage of the connectivity relationship between words, instead of using syntactic and semantic analysis. Our MT method converts the source language into the target language by permuting translated words based on the unique relationships among individual words and hence has the advantage that it uses only morphemic information. To demonstrate our method, we present results from an experimental translation system constructed on a personal computer, in which 200 sentences extracted from IEICE Transactions are translated. The fraction of sentences successfully translated was 60% for English-Japanese translation and 57% for Japanese-English translation. © 1997 Scripta Technica, Inc. Syst Comp Jpn, 28(8): 74–83, 1997