Some examples of these speech impediments are stuttering, dysarthria, cerebral palsy, brain injuries, vocal cord paralysis, and apraxia.
There are about 9 million people in the U.S. with speech issues and just about all of the speech recognition technologies are not made to help those people be understood. Todd Mozer, CEO of a company called Sensory, says that, "Speech recognizers are targeted at the vast majority of people at that center point on a bell curve. Everyone else is on the edges.” To make matters even worse, speech impediments can very much be different for every person, which would make it even harder for companies to tailor their technologies to these people.
There are several solutions out there some companies have been trying, but their focus is still not on the smaller part of the population who has these issues. In 2012, some companies began using neural networks to power their products. These networks learn from a variety of speech samples and predictable patterns. Some of these newer technologies only have about an 8% error rate.
Some other companies are looking into lip reading technologies, which would be a small fix (because some speech issues have to do with motor skills, which may make your speech unreadable) but it would nonetheless be a supplement to the technology they are already using.
As data becomes more available in the realm of speech recognition software, we can only hope that companies will implement that data to create a more inclusive environment for those of us who aren't so lucky to have fluent speech. Check out these two videos below that poke some fun at speech recognition technologies.
This tech talk was very interesting to me because this is something I don't typically think of. I think that while people with speech disabilities, or any other disabilities, shouldn't be excluded, I think it is important to realize and start improvements for these groups, such as with the lip reading technologies.
ReplyDeleteWhat other types of technologies could be implemented or created to improve this tech for people with speech disabilities?
Hey Alex!
DeleteThanks for your comments. One thing to think about is what technologies could also help people possibly speak more fluently? Of course it would be great if these companies catered to people with speech disabilities, but I like to think about how to best help the people be the best versions of themselves out in the real world. I think if companies can work together with people with speech impediments, there can be a variety of technologies that come to the market to improve voice recognition, and also improve people's ability to be more fluent.
This is very interesting topic. As this kind of technology gets used more often, it is important that it can be used everyone, including those with disabilities. While I don't know much about the technology behind this, I have to imagine that speech recognition software can be coded further to help accurately identify what someone with a speech impairment is saying. Just like speech recognition software has improved over time at recognizing different people's speech so I hope that in the future this will be advanced further to be able to better recognize what those with speech impediments are saying.
ReplyDeleteHey Rachel,
DeleteThanks for your comments. I agree with you completely, but I unfortunately think that at this time the companies who make these devices are more worried about perfecting what they already have, as well as making it available to more languages that people speak. Hopefully in the future we will see some improvement, but for now all we can do is wait and see what happens!
I never considered these types of challenges for voice recognition technologies. I guess I assumed that it was one of those technologies that work as they should. However, this tech topic made me think about my own struggles with voice recognition. I have also had trouble getting Siri to understand basic commands. This is usually attributed to my accent but it's not until I enunciate the commands that it begins to work. There are clearly still plenty of challenges for voice recognition technology and I do agree that handling speech disabilities will be one of the more difficult task to accomplish.
ReplyDeleteIt will be really intriguing to see how developers attack this issue in the future. Like you said, speech impediments can vary widely from person to person, so one simple solution likely won't solve the overarching issue. There's also the issue of serving people outside of the mean of the "bell curve" you mentioned. People from different regions speak with different inflictions or accents, which creates a similar problem for the user if the product can't accommodate to more emphasized accents. Of course, there's obviously large groups of people who experience these issues, so I'm sure it won't be hard for consumers to gain enough traction to demand solutions for these products, especially when they're becoming more and more popular in the common household.
ReplyDeleteI knew that speech recognition tools such as Siri had difficulty recognizing people with accents, but this article was very eye opening because I had never considered the difficulties that people with speech disabilities would have. Although I have to say because there are varieties of speech impediments, it probably will take a while for them to prepare one solution. I would think it would be somewhat of a continuous improvement process. But nonetheless as people become more and more aware of the needs, there will be a breakthrough to solve the issue.
ReplyDelete