Automatic Speech Recognition and Machine Translation

Abstract

The present paper deals with different aspects of automatic speech recognition. First, automatic speech recognition was carried out solely on acoustic basis. However, this approach turned out to be unsatisfactory. It seemed ressonable to complement the methodacoustic by a linguistic one. The linguistic information used in automatic speech recognition consists of different phoname statistics and on the other hand, of grammatical information of various kinds. The grammatical information allowed for in automatic speech recognition links up with some results obtained in the mechanical analysis of texts well known in machine translation. In this way, automatic speech recognition may profitably make use of the different methods developed within the investigations on mechanical analysis.