Tense and Mood Decision with Similarity Search in Japanese to Spanish Machine Translation
Tsutomu Endo · 2009
Tense and mood are pieces of information normally unattended in Japanese to Spanish machine translation as they do not exist formally in the former language. In this paper we propose a new technique to solve this issue by using a tree distance function and the nearest neighbor approach in order to find similar sentences in a knowledge base. A query sen- tence is input and it is transformed into a tree using a language model; then, a series of similar sentences is retrieved from the knowledge base; the most similar is selected and all the information of the predicates in it is assigned to the predicates in the query sentence. Our technique proves to be accurate as it obtained a 61% of correct results, just below the 63% obtained by Systran.