voice assistant It perfectly demonstrates the fact that new technology is not neutral. Such prejudices can be found in many fields. For example, the controversy surrounding an image recognition algorithm that confused blacks with gorillas is well known. Going back even further, we know that car seat belts were designed with the male anatomy in mind. Voice assistants use machine learning to understand users. Databases are often based on standard dictionaries. This means that a large proportion of people with conditions such as cerebral palsy and stuttering are often left behind when they need it most. Fortunately, big tech companies are working to improve speech recognition so that systems already exist to recognize the language of deaf people. One of his latest is Apple, where he publishes an article about working with his database of 32,000 clips sourced from podcasts.
The goal of the company founded by Steve Jobs is to enable voice assistant Siri to interpret pauses, prolongs, repetitions, and incomplete words. Based on the Podcasts database and his FluencyBank stuttering events, preliminary results show that each dataset improved his accuracy by 28% and 24%. One of the main issues with Siri so far has been interpreting stuttering pauses as the end of a sentence and returning poor results. The researchers, who published an article in arXiv, an open archive of scientific research, say the technology can also be used by people who suffer from dysarthria, or difficulty articulating phonemes due to damage to the nervous system.
A joint effort by leading technology companies
Apple isn’t the only company working towards a more comprehensive speech recognition system. First, Google is collecting more diverse audio samples to meet the needs of this segment of the population. And as part of the Euphonia project, we’re already testing a prototype app that allows people with unconventional speeches to train their devices to consider specific speaking styles.
Second, in December 2020, Amazon announced that it would integrate the Israeli startup’s technology into its Alexa assistant. Similar to Google’s project, this technology will allow each user to train the algorithm with their own characteristics. This option is expected to be operational through 2021.
Until now, voice assistants have relied on general speech patterns and tonalities beyond specific accents. However, the challenge of extending speech recognition to people who stutter or have dysarthria is believed to be much more complex. First, because the database is small, and second, because the variability of speakers is infinite. Fortunately, advances in artificial intelligence and machine learning are opening the door to a new era of accessibility in all areas. voice assistantIf you want to learn more about these types of applications, I recommend this article on how to improve accessibility using wearables and smartphones.
sauce: wall street journal
