next up previous contents index
Next: Application vocabulary and confusion Up: Speech recognition systems Previous: Branching factor (perplexity factor)

Rejection mode


It is clear from different field  trials that the application users may be involved in other conversations and are likely to elicit words out of the expected lexicon such as extra-words, synonyms, and extra-linguistic sounds. The user may be using the system in noisy  conditions and if the system is sensitive to the environment noise  it will detect a word even if the user did not speak. In order to take such phenomena into account the system has to possess rejection capabilities. 

Different approaches allow implementation of such functionality. The most common approach is a rejection model  which is trained on speech data from words that do not belong to the application vocabulary  (called trash model ). In isolated word  HMM systems, the trash model  is also an HMM  or several HMM models  beside the models of the application vocabulary .

Another approach is usually based on thresholds about the speech recognition scores (likelihood, distance measures, etc.).

If the option of rejecting out-of-lexicon words   or/and extra-linguistic phenomena is offered, then the technology provider has to explicate whether:


EAGLES SWLG SoftEdition, May 1997. Get the book...