Original article was published by TILDE on Artificial Intelligence on Medium
Speech technologies — the opportunity for people with hearing loss to be a part of public discussion in real-time
Tildes Speech technologies for the Latvian language has made more than 50 public discussions in the Conversation Festival LAMPA available to people with hearing loss. Tilde’s speech recognition technology, entirely carried out by artificial intelligence, allowed individuals to follow events with real-time subtitles.
Subtitles are an essential part of accessibility
Access to information is an essential factor in the quality of life of each individual. For people with a hearing impairment, the only possibility to access information is through sign language or subtitles; but there is the drawback — they are not always available, and are especially rare for the smaller, complex languages. Inese Immure, The Vice-President of the Latvian Association of the Deaf, explains:
“Unfortunately, the majority of cultural events are unavailable to people with hearing loss, but luckily AI-powered technologies can improve the terms for those individuals to take part in society.”
Tilde is one of the pioneers in a business environment that has taken steps to provide social inclusion and opportunities for people with hearing disabilities. Tilde’s artificial-intelligence business developer Mārtiņš Sūna explains:
“It is important for Tilde as a company, to ensure that thanks to world-class technologies, information is available to all of society, including deaf and hard of hearing people.”
Artificial intelligence technologies — a bridge between the world of sound and silence
For four days, “Tildes Balss” audio transcription, or real-time subtitles, gave the opportunity to people with a hearing impairment to follow more than 50 events and public discussions of the Conversation festival LAMPA. In total, Tilde provided at least 47 hours of real-time content with more than 383 000 words — even more than the average Game of Thrones book.
Subtitling discussions with artificial intelligence in real-time — technological challenges
Turning real-time speech into text is a technically complex process — the sound processed by the computer is translated into phonemes, and the reliability of each of the possible options is checked based on the principles of a static relationship. Afterward, the grammar algorithm divides the text into meaningful sentences, adds punctuation marks, and delivers it in a readable, enjoyable, and perceptible way. The quality of the recognized text is influenced by a variety of factors — discussion culture, terminology, foreign words, and even style of speech.
Despite the generic complexity of the technology, a speech recognition solution had to “collaborate” with third-party video streaming platforms and provide real-time titration of three parallel discussions, which means significantly more content than ever. Tilde’s artificial-intelligence business developer Mārtiņš Sūna noticed, that despite all these challenges, the technology was able to achieve the expected results:
“Technology was capable of providing sufficient quality audio transmission to the text, which was positively valued by the Latvian Association of the Deaf — people felt comfortable with this format, which is not so widely available in everyday life.”
An inclusive society driven by enterprises
The positive feedback from individuals motivates Tilde to improve speech recognition and transcription technology so that in the near future, deaf and hearing-impaired people can be fully involved in public discussions in their everyday life.
Tilde’s artificial intelligence business developer Mārtiņš is convinced that an inclusive society is essential all over the world, and enterprises can drive these changes with their contribution:
“The development of speech recognition and synthesis technologies is a slow and complex process, as there is a need for major investment; but with our example, we invite other organizations to contribute and think about the steps they can take to create a friendlier environment for everyone.”
Interested in utilizing speech technology into your workflow?
We’re here to help!