| 摘要 |
PROBLEM TO BE SOLVED: To realize a translation device which can reduce the deviation between the recitation time of speech synthesized phonetically based on a translation result and the reproduction time of an original word. SOLUTION: When the difference between the total time of utterance when a translation text is uttered by synthesized speech and the reproduction time of a speech recognition text exceeds a prescribed value (S38: No), a re-translation candidate subjected to weighting of level N1 or more is selected (S46: Yes, S48), and a synonym with the highest synonym level is selected among synonyms set as the re-translation candidate (S50: Yes, S52). Then, a selected synonym is replaced with a previous translation word, the total of the utterance time is calculated again, and it is judged whether the difference from the reproduction time is within the prescribed time (S54-S58). That is, re-translation is repeated the prescribed number of times in the extent that the meaning of the original word is not greatly changed until the difference between the total time of the utterance and the reproduction time becomes within the prescribed time. |
