Details of Grant 

EPSRC Reference:	GR/L81253/01
Title:	CONFIDENCE MEASURES FOR SPEECH RECOGNITION
Principal Investigator:	Cox, Professor S
Other Investigators:	Cawley, Dr G
Researcher Co-Investigators:
Project Partners:	BT Entropic Cambridge Research
Department:	Computing Sciences
Organisation:	University of East Anglia
Scheme:	Standard Research (Pre-FEC)
Starts:	16 February 1998	Ends:	15 August 2001	Value (£):	142,353
EPSRC Research Topic Classifications:	Human Communication in ICT
EPSRC Industrial Sector Classifications:	No relevance to Underpinning Sectors
Related Grants:
Panel History:
Summary on Grant Application Form
As the performance of automatic speech recognition (ASR) improves, it is becoming attractive to use ASR within a speech understanding system i.e. an intelligent system which accepts speech input and uses it to automatically perform a task e.g. ticket reservation, home shopping, information retrieval. These systems have the potential to bring about enormous cost savings. There will be many occasions when an ASR system will be unsure about what it has 'heard' and this uncertainty is sometimes expressed in the fact that some ASR systems produce several 'hypotheses' about what was said. However, at present, the systems have only rudimentary ways of quantifying this uncertainty and it is not clear how this should be used by the system so as to enable the most efficient completion of the task.As the performance of automatic speech recognition (ASR) improves, it is becoming attractive to use ASR within a speech understanding system i.e. an intelligent system which accepts speech input and uses it to automatically perform a task e.g. ticket reservation, home shopping, information retrieval. These systems have the potential to bring about enormous cost savings. There will be many occasions when an ASR system will be unsure about what it has 'heard' and this uncertainty is sometimes expressed in the fact that some ASR systems produce several 'hypotheses' about what was said. However, at present, the systems have only rudimentary ways of quantifying this uncertainty and it is not clear how this should be used by the system so as to enable the most efficient completion of the task. This proposal seeks to quantify the degree of uncertainty associated with the automatic decoding of a speech utterance into a 'confidence measure'. Confidence measures will be extremely useful in the development of any system which utilises ASR and in particular, in the development of speech understanding systems. Furthermore, the concept of the use of a confidence measure has a wider application to any intelligent decision-making system.
Key Findings
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Potential use in non-academic contexts
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Impacts
Description This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk Summary Date Materialised
Sectors submitted by the Researcher
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Project URL:
Further Information:
Organisation Website:	http://www.uea.ac.uk