Subfields of spoken language systems

Spoken language systems cover the following subfields (see also Gibbon et al., 1997a, User's guide, Section 1.2.1):

• automatic speech recognition: dictation software, which converts spoken utterances to written texts;
• automatic speech generation/speech synthesis systems: this covers text-to-speech systems or read-back systems that use a grapheme-phoneme conversion to produce audible and understandable utterances;
• speech input/output systems (`speech understanding' systems): systems combining automatic speech recognition and synthesis which are able to process semantic information;
• spoken dialogue systems and speech-to-speech translation systems: systems mediating, for example, between different language communities by translating spoken utterances (inclusive of speech understanding systems and at least a text-to-speech system in the target language);
• speech coding
• speech analysis or paralinguistic processing: prosodic analysis, speaker identification;
• general speech processing;
• multimodal speech recognition and synthesis systems: the most general approach of all systems mentioned combining not only speech input/output systems but also systems which use non-verbal communication such as gestures, facial expressions, and typed input (see fig  1.1).

 
Figure 1.1: Multimodal systems  

Multimodal systems  are the most powerful sets in the field of spoken language systems, it might even be considered to define the domain as multimodal systems and to see spoken language systems as a subset thereof. Due to historical reasons, (Spoken Language Systems  has a longer tradition as domain) in this paper spoken language systems will be the top domain with the subdomains, including multimodal systems, mentioned above.

Thorsten Trippel
Fri May 21 13:04:11 MET DST 1999