“Watch, Attend and Spell” utilizes machine learning algorithms to provide better IT services for persons with hearing disabilities
Watch, Attend and Spell (WAS) is a new artificial intelligence system that uses machine learning methods to read lips based on information from different audiovisual materials from six main TV programs. The software, which could be very useful to support persons with hearing disabilities using ICTs, was developed by Oxford in collaboration with DeepMind.
The WAS software tests indicated that through machine learning algorithms, the recognition of lip movement patterns was more accurate than when human experts performed this function. For this reason, the implementation of this technology may generate relevant benefits for persons with disabilities.
According to Jesal Vishnuram, Action on Hearing Loss Technology Research Manager, the development of new technology that helps people who are deaf or have a hearing disability may support them to have better access to television through superior real-time subtitling. Furthermore, he adds, technology for lip-reading may enhance the accuracy and speed of speech-to-text in noisy environments thus promoting further research.