Limitations
The findings describe this sample and these experiments. Teachers need to account for the remaining uncertainty when reviewing a result.
One-school sample
The current findings come from one school. They need validation at other schools before they can be applied more widely.
Differences between languages
Transcription errors and scoring errors differ between English and Filipino. Calibration requires a language flag, and we report results for each language.
Teacher disagreement
Teachers sometimes give different scores or reading levels to the same recording. Comparisons with automated scores must account for that disagreement.
Individual scoring error
WCPM calibration reduces the average offset, but individual scores can still differ substantially from a teacher’s. Accuracy calibration provides too little information for individual display.
Unvalidated word highlighting
Highlights help teachers compare a transcript with the passage. The study does not validate them as word-level mispronunciation detection.
Teacher-assigned reading levels
The teacher assigns the Phil-IRI reading level and makes the instructional decision. READSmart does not determine a final level automatically.