A phoneme is the smallest unit of sound that distinguishes one word from another in a language. The only thing separating "bat" from "pat" is a single phoneme, the sound at the front. Speech recognition has historically modeled speech as sequences of phonemes, mapping audio to these units first and then assembling them into words.
How speech systems use phonemes
Every language has its own inventory of phonemes, and English has roughly 44, depending on accent and how you count. A pronunciation lexicon maps each word to its phoneme sequence, so "cat" resolves to the three sounds that make it up.
Classical recognizers used an acoustic model to identify phonemes from the audio, then combined those phonemes into words with the lexicon and a language model. That pipeline is why pronunciation dictionaries were once central to building any speech system.
Phoneme, phone, and grapheme
Phoneme, phone, and grapheme are three similar terms that cause most of the confusion. A phoneme is an abstract, contrastive unit specific to one language. A phone is an actual spoken sound as produced, and the real-world variants of a single phoneme are called allophones.
A grapheme is a written letter. Because letters and sounds line up poorly in English, converting spelling to pronunciation is its own task, grapheme-to-phoneme conversion, needed precisely because English spelling is an unreliable guide to how a word sounds.
Why phonemes matter
Phonemes underpin pronunciation lexicons, custom pronunciations, and forced alignment. Modern end-to-end models often skip explicit phonemes and predict characters or subword units straight from audio, so the concept is less mechanically central than it once was.
Phonemes still explain a lot in practice, though. The concept clarifies how a system handles unfamiliar names and rare words, why two spellings can sound identical, and why some providers let you supply a custom pronunciation when the default reading of a term is wrong.
Frequently asked questions
What is a phoneme? It is the smallest unit of sound that changes meaning in a language. Swap one phoneme and you get a different word, as in "bat" versus "pat" or "cat" versus "cut."
What is an example of a phoneme? The /b/ at the start of "bat" is a phoneme, as is the /p/ that replaces it in "pat." English has about 44 such sounds across its vowels and consonants.
What is the difference between a phoneme and a grapheme? A phoneme is a unit of sound; a grapheme is a unit of writing, a letter or letter group. The word "ship" has four letters but three phonemes, since "sh" spells one sound.
Do modern speech recognition systems still use phonemes? Many end-to-end models predict characters or subwords directly and skip explicit phonemes. Phonemes still power pronunciation lexicons, custom pronunciations, and alignment tools, so the concept remains useful.