AI Tool Transcribes Lectures for Students, Aiding Accessibility
Translated & summarized from Israel Hayom by baba
Students at Tel Aviv-Yafo Academic College have developed an AI system called Accessible Academic that transcribes Hebrew lectures into text, generating subtitles, summaries, and study flashcards. The tool, created by hearing-impaired student Adi Tefiro and Matan Gurfinkel, aims to improve accessibility for students with hearing loss but can also help reservists and new immigrants. The system processes an hour-long lecture in 7-12 minutes and can convert spoken math into written formulas. It is currently in use by hearing-impaired students and is planned for wider release, likely for a fee.
The story in 6 lines · by baba
- An AI system named Accessible Academic transcribes Hebrew lectures into text, generating subtitles, summaries, and flashcards.
- The tool was developed by students Adi Tefiro and Matan Gurfinkel at Tel Aviv-Yafo Academic College.
- It aims to assist students with hearing impairments but also benefits reservists and new immigrants.
- The system processes an hour-long lecture in 7 to 12 minutes.
- It can convert spoken mathematical content into written formulas within study materials.
- The system is currently used by hearing-impaired students and will likely be offered commercially.
A new AI-powered system called Accessible Academic has been developed to transcribe Hebrew academic lectures into text and create comprehensive study materials. The system generates word-level synchronized subtitles for videos and simultaneously produces summaries and flashcards. Initially designed to assist deaf and hearing-impaired students, the tool can also benefit reservists who missed classes, new immigrants, and any student preferring written material over watching full lectures.
The system was created as a final project by Adi Tefiro, a student with a hearing impairment, and Matan Gurfinkel, both from Tel Aviv-Yafo Academic College's School of Computer Science and Artificial Intelligence. After uploading a lecture recording, the AI provides a full Hebrew transcription with precise timestamps, identifies key topics, and generates structured summaries and practice questions. Processing an hour-long lecture takes approximately 7 to 12 minutes. All generated materials undergo a four-stage review to ensure accuracy and prevent external content inclusion. The platform allows users to read the full lecture, quickly jump to specific points in the recording, or study using the generated aids. Notably, it can convert spoken mathematical content into written formulas within summaries and flashcards, a feature not currently available in other transcription tools.
Tefiro, who has been hearing impaired since age 14, was motivated by her personal struggles with lectures, including poor acoustics, uncooperative lecturers, and the inadequacy of automatic subtitles for complex subjects like linear algebra. "When the final project came, it was clear to me what I wanted to build," she stated. Gurfinkel readily joined the project, and together they developed the platform. They realized the system's broader applicability, noting, "It's good for everyone, for the student with a learning disability, for the reservist who returned after two months, for the mother who was watching the baby. Accessibility is not convenience. It is a commitment. Everyone deserves to hear a lecture recording."
The system was developed under the supervision of Dror Margalit and was presented at the School of Computer Science and Artificial Intelligence's final project competition. The system is currently used by students with hearing impairments and is expected to be released to a wider audience, likely on a paid basis.