[Articles] The state of speech recognition

[CT Droid]

Found this very interesting article regarding the history and state of the speech recognition industry.  Make sure to read the comments as well, since they contain some interesting feedback as well from industry insiders.

Rest in Peas: The Unrecognized Death of Speech Recognition



View the full article

 

[Dave X10]

Found this very interesting article regarding the history and state of the speech recognition industry.  Make sure to read the comments as well, since they contain some interesting feedback as well from industry insiders.
Rest in Peas: The Unrecognized Death of Speech Recognition
View the full article

Thanks for sharing! I use BVC to control much of my automation. I don't really have many problems.... but I have taken care to select words/phrases that don't sound too much alike. I am very happy with my Voice Control.

 

[Lagerhead]

Related to this topic is Hubert Dreyfus' 1972 book, "What Computers Can't Do" -- an early and, in retrospect, entirely accurate assessment of the limits of computation. At the time Dreyfus was mercilessly derided by the Artificial Intelligence establishment. I also have a laughable Bell Labs monograph from 1959 that, based on some early success in proving mathematical theorems, claimed that the major mathematical puzzles would eventually be solved through automated investigation.

 

[Dave X10]

Related to this topic is Hubert Dreyfus' 1972 book, "What Computers Can't Do" -- an early and, in retrospect, entirely accurate assessment of the limits of computation. At the time Dreyfus was mercilessly derided by the Artificial Intelligence establishment.

Certainly... all machines (and biological organisms) will always have limits. However... I would love to see a Web Based interactive chat-bot that could be intergraded into my Home Automation system. Right now my Voice Commanded Setup... merely responds to X10 PLC and RF signals or my Voice Commands. Some level... even a tiny little bit of AI... could really help my setup seem to come to life.

 

[electron]

Unless you are single, I can tell you for sure it will get old fast (I used to specialize in clever chat bots)

 

[Dave X10]

Unless you are single, I can tell you for sure it will get old fast (I used to specialize in clever chat bots)

Well... then you may be the person I should share my idea with. I envision a chat bot that can be "tuned" to use the Internet to check RSS feeds for information that would be of interest to the user.

I currently use WeatherAloud with my BVC setup. So I can ask for "temperature" and get the current weather conditions. I have also created a "automated chair" macro... that determines (if) that I just got up... and tells me the current weather as it turns on my TV. It would be nice... If a program/chat bot could look over RSS feeds and maybe give me a heads up on any "breaking news" or weather alerts (or other preselected information) in bite-sized amounts. I could use NewsAloud... but I really want my info screened or reduced to fewer words. This sort of "news in few words" may actually require a Web Site designed and dedicated to a chat bot/automation type program application.

It would be actually more human like if the bot could report or recite.. or discuss less important information.. on request or by condition. Like what DVDs will be released on Tuesday, if a friend wrote on my wall (in facebook), if my son emailed me, or... whats on TV tonight. I've been working (in a very loose use of the word work) or a project I call S.A.R.A. (Smart Articulate Responsive Automation) in an attempt to expand the computers role as an automation device.

I could be wrong... but I think the PC program/Internet enabled chat bot would be the perfect automation interface.

 

[damage]

there's this HAL app: http://www.james.lipsit.com/tell_me.htm but that means you'd have to run HAL with all its point and clunk glory. it does get old fast. my kids don't even care to ask the house things anymore. they'd rather play with a different tell me on my cellphone http://www.tellme.com/you - because they like seeing how badly it can mangle what they said

 

[sic0048]

Well as someone that uses the translation feature of Google Voice, I can definitely say that voice recognition is not very good. That may be the understatement of the decade too. Many times, I have not idea who called or what they wanted from looking at the translation - it's that bad. Perhaps 20% of the time, I can tell who called, but rarely, if ever do I understand the message without actually listening to it.

 

[damage]

Well as someone that uses the translation feature of Google Voice, I can definitely say that voice recognition is not very good. That may be the understatement of the decade too. Many times, I have not idea who called or what they wanted from looking at the translation - it's that bad. Perhaps 20% of the time, I can tell who called, but rarely, if ever do I understand the message without actually listening to it.

yeah it's pretty much a joke. the voice search feature of google maps mobile works decently though.

 

[Dave X10]

Well as someone that uses the translation feature of Google Voice, I can definitely say that voice recognition is not very good. That may be the understatement of the decade too.

And from the article [that started this thread] it appears that Speech Recognition won't be improving [at least in any leaps or bounds] in the very near future. But my system... and I would guess most others... does perfectly well with yes and no input/responses. So whereas... asking the computer to perform advanced calculations [like in a Star Trek episode] isn't realistic... Saying yes to a suggested wake up call is easy enough to do. And it works. Getting the system to ask at the right time is a [macro] function of timers and occupancy sensing.

Currently I am not looking at improving the Speech Recognition (I wouldn't know how... or even where to start). I am looking at using what technologies that do work well and exploiting them [maybe in new ways] to get more use out of the speech recognition/Home Automation setup.

Jame Lipsit has done much fine work... and has accomplished a great deal. I admire what he has done with [and for] voice controlled automation. But I don't think I will be pursuing HAL.

Although it can be nice to have a little "show" of sorts to demonstrate my BVC Home Automation. Living with the system has me speaking to it very little... but listening to Kate (the Kate 16 voice) for warnings, alerts, and even the time.