phoneme labelling

Terms from Artificial Intelligence: humans at the heart of algorithms

The glossary is being gradually proof checked, but currently has many typos and misspellings.

Phonemes are small segments of speech sound such as the 'aye' sound in AI. Phoneme labelling is the stage within speech recognition of taking individual phonemes in a speech source and labelling them with the appropriate phoneme tag. Typically for vowels this is largely about the frequency distrubution, but for consonants the shape of the sound is important, particularly the 'attack', the way the sound starts. This is made more complex as it can be hard to separate the phonemes in the audio stream and also because the precise qualities of one phoneme often vary depending on the immediately preceding one. Phoneme labelling was one of the early uses of Kohonen networks.

Used in Chap. 6: page 84