Speech recognition and synthesis

Speech recognition software attempts to decipher the spoken language into form that computers understands. Core part of this process is signal processing and thus heavily depends on mathematical formulas and such, but in the later stages where the software already has vague idea of what is the textual representation of the spoken input is, introducing some linguistic information to processing may help to identify the words and sentence structure of the input.

The opposite direction, producing synthesised speech from textual input, can likewise benefit from the linguistic information as the pronounciation of words is related to their part-of-speech, position in sentence and similar features.

The NLP APIs available on this site currently do not provide any tools for spoken language, but as mentioned earlier the linguistic analyser APIs can be utilized as part of speech recognizer or synthesizer applications.