This paper presents an architecture and a prototype for speech-to-speech translation on Android devices, based on GF (Grammatical Framework). From the user's point of view, the advantage is that the system works off-line and yet has a lean size; it also gives, as a bonus, grammatical information useful for language learners. From the developer's point of view, the advantage is the open architecture that permits the customization of the system to new languages and for special purposes. Thus the architecture can be used for controlled-language-like translators that deliver very high quality, which is the traditional strength of GF. However, this paper focuses on a general-purpose system that allows arbitrary input. It covers eight languages.
This paper describes how grammar-based language models for speech recognition systems can be generated from Grammatical Framework (GF) grammars. Context-free grammars and finite-state models can be generated in several formats: GSL, SRGS, JSGF, and HTK SLF. In addition, semantic interpretation code can be embedded in the generated context-free grammars. This enables rapid development of portable, multilingual and easily modifiable speech recognition applications.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.