How does sayit score word stress and intonation?

Word stress and intonation are scored separately from the individual sounds in your speech, because you can pronounce every phoneme in a word correctly and still be misheard if the stress lands on the wrong syllable, or sound flat and hard to follow if your pitch never moves. sayit measures both from the same recording your pronunciation is scored from, using loudness, duration, and pitch rather than the words alone.

30-second version: For multi-syllable words, sayit compares the loudness and length of each syllable against where the stress is supposed to fall, and reports what fraction of measurable words got it right. Separately, it tracks your pitch across the whole sentence and checks whether it moves the way a statement, question, or list is expected to — rising, falling, or flat — against the pattern for that sentence type.

How does sayit know where a word's stress should be?

From the same phonetic source everything else in the pipeline uses: the espeak phonemizer marks the stressed syllable when it converts a word to IPA. sayit compares the energy and duration of each syllable in your recording against that expected pattern — the stressed syllable in a correctly-stressed word should stand out as louder and a little longer than its neighbors. Say comFORtable when the word should be COMfortable and the mismatch is measurable directly from the audio, independent of whether each individual sound was pronounced cleanly.

Why does stress matter as much as the sounds themselves?

Because English listeners use the stressed syllable as a key part of how they recognize a word at all. Move the stress and a listener can fail to recognize a word they know perfectly well, even when every phoneme in it was correct — which is a different, and often more damaging, failure than a single mispronounced sound, because the listener doesn't get a garbled word to puzzle over; they get a word that doesn't match anything in their mental dictionary.

What does the intonation score measure?

Your pitch contour across a whole sentence — whether it falls at the end of a statement, rises for a genuine yes/no question, or stays essentially flat when it should be doing one of those things. sayit extracts your fundamental frequency over the length of the sentence and compares its overall shape to the pattern expected for that sentence type. A flat delivery on a question you meant to ask, or a falling tone where a rise was expected, shows up here even when every word in the sentence scored perfectly on its own.

Why is my stress score sometimes missing instead of a number?

Because stress can only be measured on words with more than one syllable, and it needs enough of them in a take to be a meaningful reading rather than a guess from one or two data points. Below that minimum, sayit reports the stress score as absent rather than publishing a number built on too little — the same honesty rule behind why a very short take doesn't get a headline score at all. When there's enough to measure, sayit also reports coverage: what share of the eligible multi-syllable words it could actually score.

Stress and intonation, side by side

Try it

Read a sentence with a real question mark and a sentence with a comfortable multi-syllable word in sayit — the results will show a stress percentage and an intonation reading alongside your accuracy, not folded into one number. The pitch contour article goes deeper into how the intonation curve itself is drawn and read.