A system that determines foreign language proficiency by tracking gaze during reading has useful application to educational testing services and to language learners around the world. This alternative approach can serve not only as a language evaluation tool but also as a language learning and assessment tool.
Traditional language testing methods require costly and time consuming efforts to regularly compose, pilot, and update test editions. Furthermore, language testing services rely on handcrafted questions and standardized formats that leave room for human error in test composition and cheating in test administration. These characteristics limit effectiveness and reliability of tests designed to evaluate foreign language proficiency.
This technology provides an innovative test to determine language proficiency by tracking gaze as English as a Second Language (ESL) learners read free-form English text. Analysis of gaze patterns enables direct measurement of cognitive linguistic ability and evaluation of language comprehension from reading. Direct measurement of foreign language proficiency in this manner can improve test reliability, reduce test development costs, and strengthen robustness to cheating as compared to existing English tests.
The developed system determines language proficiency by tracking the gaze of ESL test takers. To determine proficiency, the system relies on an eye tracker camera that records gaze patterns of test takers as they read free-form English sentences and subsequent comprehension questions.
The technique includes a machine learning algorithm to process collected gaze patterns and to compare patterns of native and nonnative English speakers. This feature set and comparison establish a baseline for how the system determines English proficiency.
Proficiency scores obtained with the eye tracking test strongly correlate with scores from standardized tests such as the Michigan English Test (MET) and TOEFL. Comparison of eye tracking test results to results from standardized English tests can serve as a test diagnosis and calibration tool as well as a tool for outcome prediction on standardized tests. Studies demonstrate fidelity in the described approach and high consistency for repeat test takers.
- Alternative approach to standardized English testing
- Language independent, potential deployment for other languages besides English