The UB will develop a real-time digital transcription system to make university teaching more accessible in Catalan
The University of Barcelona (UB) will develop an innovative real-time digital transcription system to make university teaching in Catalan more accessible. The system is based on artificial intelligence (AI), is called SCRIBAL and is specially adapted to the different dialects of Catalan and the specific terminology of universities. The project, led by Mireia Farrús, professor of the Department of Catalan Philology and General Linguistics of the Faculty of Philology and Communication of the UB, has received a grant from the Industry of Knowledge program of the Department of Research and Universities of the Generalitat de Catalunya, to develop a prototype and validate it in the classrooms of the University.
The technology makes it possible to quickly and accurately transcribe classes, lectures and other academic content into audio, facilitating study and improving accessibility for students with hearing problems or who are not fluent in the language of instruction. “Most educational institutions do not offer an audio-to-text transcription service due to technological limitations and the costs of manual transcription. With this digital system, students with hearing disabilities will be able to fully access classes and lectures, enhance their learning experience and reduce the barriers that often prevent them from making the most of their time at university.”
SCRIBAL is the only project in the area of Humanities that has received a grant in the Product modality in this call. This aid is aimed at obtaining prototypes and the valorization and transfer and is part of the Knowledge Industry program, an initiative of the Department of Research and Universities of the Generalitat de Catalunya to encourage the transfer of research results.
Integrated solution for educational content management
Based on advances in speech recognition and natural language processing technologies, SCRIBAL is a system that allows the transcription of the teacher’s oral content through the subtitling of PowerPoint presentations or directly on the student’s computer or cell phone.
Another innovative aspect of SCRIBAL is its ability to identify different speakers during a class or conference, synchronize the text with the audio and export the transcripts to various formats, thus facilitating their integration with e-learning platforms. In addition, students who are not fluent in the language of instruction will be able to use the transcripts to translate the content into their native language, improving their comprehension and academic performance. “These functions make SCRIBAL not only a transcription tool, but also a comprehensive solution for the management and distribution of educational content,” explains the researcher.
The system also offers significant advantages in terms of privacy and security. “As it is self-managed by the university, student data and class content can be stored and transmitted securely, complying with privacy and data protection regulations,” emphasizes Mireia Farrús.
A system optimized for Catalan and with a gender perspective
This new solution is based on OpenAI’s open source transcription system Whisper. It is a neural network technology – a field of artificial intelligence that uses computational models inspired by the human brain to recognize patterns and solve complex problems – that is at the forefront of speech recognition and has been trained with hundreds of thousands of hours of speech.
The researchers have incorporated the CommonVoice and ParlamentParla databases into the system to adapt it to the different varieties of Catalan. The result has been an increase in the accuracy of the transcriptions with a significant reduction in the error rate per word. In addition, these databases have been carefully balanced to take into account the gender perspective in their development, thus avoiding the biases of other systems.
The project, led by a multidisciplinary team of artificial intelligence researchers and entrepreneurs, is at an advanced stage of technological maturity, validating its technical feasibility in relevant environments. Now, with this 2024 PROD 00007 grant of 150,000 euros, the goal is to develop a complete prototype with user interface that can be used in real time in classrooms. In addition, pilot tests will be carried out at the UB with users to validate the acceptance and usefulness of the solution.