APPAREIL ET PROCEDES DE RECONNAISSANCE VOCALE

摘要

The present invention relates to a speech recognition apparatus for transcribing speech and adapted to reduce the probability of rendering an incorrect homophone from a set thereof. Selected words in the vocabulary of the apparatus have associated therewith respective tag values or states characterizing at least one of the following parameters thereof (i) number (e.g. singular or plural); (ii) gender (e.g. masculine, feminine or neuter); (iii) person (e.g. first, second or third); and (iv) any other characteristic of the word which determines the form of a following word or which clarifies the selection of an appropriate homophone from a set thereof. The selection of a particular homophone in transcribed speech is based upon the tag values or states of one or more previously transcribed words.