If strings such as '3+1=0' or '3 > 2' come out as provable in the system (arithmetic modulo 4, say) then that is enough to count them as correct utterances of the system.
On the other hand, transcription errors may introduce mistakes and lead to misses of correct utterances, especially on large queries: the longer the searched term is, the higher the probability of encountering an erroneous word is.
But insofar as they are capable of explicating the conditions for valid or correct utterances, they also explain why some utterances are invalid, some speech acts unsuccessful, and some argumentation inadequate.
Matches were considered to be 'correct' when utterances were associated with the correct gestures, and utterances without any accompanying gestures were considered 'correct' when no gestures were associated with them.
Kaplan believes that if this ambiguity is eliminated too soon, the correct meaning of an utterance-especially a long or complex sentence-may be lost.
In supervised mode, the correct transcription of the adapting speech utterance is available.
It is interesting to note that the rescoring process is especially effective in correcting utterances with errors in large vocabulary continuous speech recognition, where constraints imposed by the lexical and language models sometimes produce recognition results not strictly observing the underlying acoustic-phonetic properties.
Instead, the referent most suitable to the utterance situation will usually be the correct one.
The utterance error rate (UER) of our speech decoder on the set of decoding results where the correct transcription was present in the LM was 10.0% (see Section 4.2).
While in this experiment the correct transcription of the response was always in the language model, our system must also be able to reject utterances when they are not present in the language model, while still accepting correctly recognized utterances.
