Results 1 to 10 of 10

Thread: Computer voice recognition vs. Alexa's

  1. #1

    Computer voice recognition vs. Alexa's

    I've been using voice recognition since the late 1990s, and over the years spent hundreds (thousands?) on software, microphones, computers. My accuracy has always been, meh. I read about people getting 98% accuracy and dictating hundreds or even thousands of words per hour. For me, a great day has always been to wrestle down a couple thousand over an entire day. Still, voice recognition kept me productive.


    Then a couple of years ago I unboxed my first Echo Dot. It cost $40. I've read it has an array of seven microphones, all obviously worth less than $40. And I don't have to tell anyone here that the voice recognition is amazing. Over loud music, over cross-talk, even into the next room, even with my sloppy diction, it understands what I'm saying.


    Then I go back to my computer with its voice recognition and desktop mic, which cost way more than $40, and it's like going from the hare back to the tortoise. I wonder, what do you think accounts for this discrepancy? I'm always thinking, "I'll take the far superior voice recognition for $40, Alex."

  2. #2
    Right now, Nuance the makers of Dragon NaturallySpeaking do not have any real competition. Nuance does not care to make their product any better. It is quite vexing for the end-user! I also own 2 Amazon echoes and 1 Google home.
    The test of success is not what you do when you are on top. Success is how high you bounce when you hit the bottom
    --General George Patton

    Complex problems need to be solved collectively.
    ––Paul Nussbaum
    usc87.blogspot.com

  3. #3
    Quote Originally Posted by Spitzbub View Post
    I've been using voice recognition since the late 1990s, and over the years spent hundreds (thousands?) on software, microphones, computers. My accuracy has always been, meh. I read about people getting 98% accuracy and dictating hundreds or even thousands of words per hour. For me, a great day has always been to wrestle down a couple thousand over an entire day. Still, voice recognition kept me productive.


    Then a couple of years ago I unboxed my first Echo Dot. It cost $40. I've read it has an array of seven microphones, all obviously worth less than $40. And I don't have to tell anyone here that the voice recognition is amazing. Over loud music, over cross-talk, even into the next room, even with my sloppy diction, it understands what I'm saying.


    Then I go back to my computer with its voice recognition and desktop mic, which cost way more than $40, and it's like going from the hare back to the tortoise. I wonder, what do you think accounts for this discrepancy? I'm always thinking, "I'll take the far superior voice recognition for $40, Alex."
    Cut the wire to The Internet and see how "good" Alexa's (et al.) performance becomes.

    Additionally, Alexa operates in a different domain. It doesn't have to deal with the possibility that you might want to utter "specious albino pontificators bifurcating merrily" -- which I would have uttered had I dictated this post to Dragon. These devices know that the sorts of utterances they will have to process are either queries (Who/What/Where/etc.) or directives (<verb> ...). They don't have to deal with arbitrary sentences ("The ball is green") or even nonsense word lists ("Portable run jump fly green toosie roll")

    You can dramatically increase recognition accuracy/speed if you constrain context. E.g., "Call Philomena" stands a much greater chance of being recognized correctly when the notion of a "telephone" and "address book" represents a typical context (even if you're unsure that it is the correct context for that utterance). Add to that the presence of an entry in the address book for someone/something having a name sounding similar to "filomena" and its a lead pipe cinch!

  4. #4
    Senior Member
    Join Date
    Aug 2001
    Location
    Melbourne, FL USA
    Posts
    1,572
    IMHO voice recognition has really come into its own this decade. Was a beta test for several voice recognition software, IBM ViaVoice and don't remember the other one (not DNS) during the 80s for about 20 years.

    During this time, two significant things occurred, the platform required became cheaper with the proliferation of quad core processors and cheap RAM. The other was the algorithm to detect what you were trying to say changed from voice recognition to voice prediction.

    Have a Dell laptop which has a built-in microphone to work with Windows 10 and configured the computer to use that for DNS 13 and it works fine. You talk to your phone, your computer, your house keyboards are going to be obsolete and the next thing down the pipe will be holographic communications.

    Voice recognition is going to be a standard for any devices requiring human input, cannot imagine what this world will be like in 20 years

  5. #5
    Quote Originally Posted by Cris View Post
    Voice recognition is going to be a standard for any devices requiring human input, cannot imagine what this world will be like in 20 years
    There is a fixed minimum requirement for resources (memory, cpu, etc.) below which any sort of meaningful voice interaction will simply be unacceptable. At the same time, there will be an ever increasing set of gadgets -- of ever decreasing capabilities -- that require "human interaction".

    It's unlikely that you'll be talking to your microwave oven, thermostat, irrigation controller, (de)humidifier, sewing machine, refrigerator, garage door opener, doorbell, oven, washer/dryer, water softener, ... (you get the picture). These devices have very slim margins and lots of pricing pressure to keep their recurring (i.e., parts+labor) costs down. You will see attempts to tie them to "centralized controllers" that can distribute the costs of the voice interface over a greater number of devices. But, the devices will still need to be able to be operated in the absence of such controllers.

    There are many applications that are simply not suited to voice control (imagine trying to drive your chair with JUST your voice! think about it!!).

    There are folks who can't speak (so, some alternative must be available for them).

    And, there are domains in which speech is not welcome (imagine a classroom full of students all trying to "dictate" their notes to their "notebooks" or a soldier deployed on a battlefield) or comes at too high of a cost (e.g., privacy).

    Speech will just be yet another user interface modality that will be evaluated with all the others that are -- or are becoming -- available.

  6. #6
    sanatino
    Guest
    Quote Originally Posted by Spitzbub View Post
    I've been using voice recognition since the late 1990s, and over the years spent hundreds (thousands?) on software, microphones, computers. My accuracy has always been, meh. I read about people getting 98% accuracy and dictating hundreds or even thousands of words per hour. For me, a great day has always been to wrestle down a couple thousand over an entire day. Still, voice recognition kept me productive.


    Then a couple of years ago I unboxed my first Echo Dot. It cost $40. I've read it has an array of seven microphones, all obviously worth less than $40. And I don't have to tell anyone here that the voice recognition is amazing. Over loud music, over cross-talk, even into the next room, even with my sloppy diction, it understands what I'm saying.


    Then I go back to my computer with its voice recognition and desktop mic, which cost way more than $40, and it's like going from the hare back to the tortoise. I wonder, what do you think accounts for this discrepancy? I'm always thinking, "I'll take the far superior voice recognition for $40, Alex."

    I want to offer you a fantastic tool that you can do with voice recognition and face recognition. Meanwhile, objects in the Internet projects are also very painful in your projects. It's a super-rescue tool that works with the Python programming language. . I propose to scold you about it

  7. #7

    re: computer voice recognition versus Alexa's

    Quote Originally Posted by sanatino View Post
    I want to offer you a fantastic tool that you can do with voice recognition and face recognition. Meanwhile, objects in the Internet projects are also very painful in your projects. It's a super-rescue tool that works with the Python programming language. . I propose to scold you about it
    Thanks, I am interested. Please tell me more.

  8. #8
    Thanks for the informative answers, I learned something. It's a huge improvement for me switching back to Dragon. When Windows 7 came out I switched over to Windows Speech Recognition because it was a seamless part of Windows (no more software conflicts or lost profiles) and the HP laptop's built-in web cam mic eliminated the need for external mics. Most importantly, in those days WSR and Dragon ran virtually neck and neck in accuracy and ease-of-use.

    Fast forward to this summer and a slightly faster HP laptop, and WSR in Windows 10 is horrendous. There has been almost no change or improvement in WSR since then (not since Vista, I've read). In fact its performance is for some reason a lot worse than the Windows 7 laptop's. No pride of product at Microsoft. Anyway, this longtime Dragon user has returned to the fold.

    Still, I will be watching for any document dictation applications from Alexa.

  9. #9
    Senior Member
    Join Date
    May 2006
    Location
    Central NJ
    Posts
    2,475
    Blog Entries
    1
    My computer guy (whose name is Brad, and is worthy of the same reverence ask her first Subaru) tells me that the voice recognition built into Windows 10 is very effective at navigation, mouse control, so that I will need Dragon only for dictation. Is that others’ experience? I’ve been stalwartly clinging to Windows 7.

  10. #10
    Quote Originally Posted by Random View Post
    I?ve been stalwartly clinging to Windows 7.
    The fact that it is almost impossible to avoid updates to a W10 machine is my biggest argument against adopting it. (YMMV)

    I run W7 on my windows machines and only install updates occasionally (typically, when I rebuild a machine or build a new one). The idea that the machine that I use tomorrow may not be the machine that it was when I went to bed the night before is unsettling (too many horror stories of their Windows machines "breaking" due to an update)

    OTOH, most of my machines are air-gapped so "security patches" aren't critical (someone would have to break into the house to "compromise" the machines -- and, at that point, they can just as easily STEAL them!).

Similar Threads

  1. Replies: 9
    Last Post: 07-12-2009, 02:02 AM
  2. best voice recognition computer software?
    By wheel123 in forum Computers
    Replies: 5
    Last Post: 03-02-2007, 10:51 AM
  3. best voice recognition computer software?
    By wheel123 in forum Equipment
    Replies: 2
    Last Post: 02-26-2007, 02:46 PM
  4. Replies: 3
    Last Post: 04-20-2006, 09:11 PM
  5. Replies: 1
    Last Post: 08-08-2005, 10:30 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •