Verification Evaluation

In an educational setting, accuracy is paramount.

An AI system for assessment that tells a student they are wrong when they are right (false negative) damages confidence and is frustrating and annoying to students and educators.
Equally a system that responds to a student telling them they are right when they are actually wrong (false positive), fails to recognize difficulties and therefore can neither assess nor assist the student as required.

Our evaluations, continuously performed, return statistically significant analysis, and are challenging in order to test the full rigor of the system and ensure the results are indicative of that system’s behavior in real-world conditions.

We rigorously evaluate our system for accuracy on a wide variety of datasets which include:

• accents and dialects

• correct and incorrect utterances

• challenging young children’s speech, not just older mature speakers

• Real world noisy speech data

• Wide variety of devices