Towards a computer that can recognize everyone's conversations
- Research on multi-speaker conversational speech recognition: past, present and future -
Takaaki Hori, Media Information Laboratory
This presentation introduces the technology we have developed at NTT Communication Science Laboratories for automatically recognizing conversational speech in a meeting situation. In recent years, smart phones have appeared that can be operated by voice commands, and opportunities are increasing for ordinary people to use speech recognition technology. However, when we try to make the device recognize actual conversations, as many as half of the words are not be recognized correctly. Why does the system not work well? In this presentation, I will talk about why conversational speech recognition is difficult for computers, what problems we should solve, and our achievements in this research area. I would also like to talk about the future we can expect through progress in this technology.