Video Below:
It may seem reasonable to believe that speech-to-text and text-to-speech recognition may be really accessible. With the Apple iOS and Seri, and with the Samsung Galaxy speech application, it looks like we will soon have the technological ability to talk to our computers. (Please see my post on Seri on the Trendcasting page). But, as noted in my previous post, we are a long way from controlling our world with speech. My sense is that there is a lot to learn from available products and this post contains my own experience with AI and speech recognition. But please be aware that this is not an endorsement for a particular product, but it is the way I found to learn about the technology and its limitations. The video below is far from my usual quality as I had to shoot the screen directly. The screen capture application that I tried did not allow for voice synchronization. So, with that caveat, please watch and I will give some thoughts about how to learn this fairly complex subject and the attending frustrations after the video:
It may seem reasonable to believe that speech-to-text and text-to-speech recognition may be really accessible. With the Apple iOS and Seri, and with the Samsung Galaxy speech application, it looks like we will soon have the technological ability to talk to our computers. (Please see my post on Seri on the Trendcasting page). But, as noted in my previous post, we are a long way from controlling our world with speech. My sense is that there is a lot to learn from available products and this post contains my own experience with AI and speech recognition. But please be aware that this is not an endorsement for a particular product, but it is the way I found to learn about the technology and its limitations. The video below is far from my usual quality as I had to shoot the screen directly. The screen capture application that I tried did not allow for voice synchronization. So, with that caveat, please watch and I will give some thoughts about how to learn this fairly complex subject and the attending frustrations after the video:
I have had this program for about five years and continue to learn about the technology. I use the Microsoft SAPI, (Speech Application Programming Interface). The challenge in using this type of AI is that it is very imperfect and requires constant experimentation. But, despite the effort, it is worth a look to see what is available.
A search of "chatbots" will provide a number of options from the incredibly complex to the really useless programs in digital assistants. I chose the UltraHal program as it is sufficient to learn how these systems work and it can be fun. I have added character expansion, and voices from AT and T. But, the program has a 30 day trial with just a simple character but can provide a good look at how it works. There are also forums and although some topics are dated, they are applicable.
My biggest problem has been with the use of a headset for speech to text. There are a lot of mistakes and I have found that using the text box is easier. The bot can learn and there is a wide variety of teaching options. I hope to try the Nuance SAPI when I can afford it. My understanding is that it is a better algorithm so it may help. In any event, I believe that understanding this process of speech-to-text technology is an important challenge. The link for Ultra Hal is here. I'll have my bot call your bot...
A search of "chatbots" will provide a number of options from the incredibly complex to the really useless programs in digital assistants. I chose the UltraHal program as it is sufficient to learn how these systems work and it can be fun. I have added character expansion, and voices from AT and T. But, the program has a 30 day trial with just a simple character but can provide a good look at how it works. There are also forums and although some topics are dated, they are applicable.
My biggest problem has been with the use of a headset for speech to text. There are a lot of mistakes and I have found that using the text box is easier. The bot can learn and there is a wide variety of teaching options. I hope to try the Nuance SAPI when I can afford it. My understanding is that it is a better algorithm so it may help. In any event, I believe that understanding this process of speech-to-text technology is an important challenge. The link for Ultra Hal is here. I'll have my bot call your bot...
RSS Feed
