Speech recognition for under-resourced languages: Data sharing in hidden Markov model systems

摘要

Speech interfaces to different types of technology are becoming increasingly more common. Users can use theirvoice to search the Internet, control the volume of their car radio or dictate. However, this possibility is onlyavailable to users if the required technology exists in the language they speak. Automatic speech recognition(ASR) technology already exists and is regularly used by speakers of American English, British English, German,Japanese, etc. The development of ASR systems requires substantial amounts of speech and text data. Whilesuch resources are readily available for a number of languages, the majority of the languages that are spokenin the world can be classified as under-resourced, i.e. the resources required to create technologies like ASR donot exist or exist only to a limited degree. Researchers in the field of speech technology development for under-resourced languages are investigating various possibilities to address this challenge and to establish resources andtechnologies in as many languages as possible.