Watson speech to text - Alfa Romeo Forum
You are currently unregistered, register for more features.    
The Technology Section A place to discuss technology & gadgets.

 
Thread Tools
(Post Link) post #1 of 21 Old 02-05-16 Thread Starter
Status: -
AO Gold Member
 
Join Date: Nov 2002
Location: Canada
Posts: 6,209

Member car:

1973 Spider 2000

Watson speech to text

Has anybody here tried this? And even if you haven't, how complicated is this actually?
https://github.com/watson-developer-...to-text-nodejs

I'm not a programmer or a developer, but I'm a pretty smart end user. I can follow instructions, and think my way through things. I'm seeing the step-by-step guide, but I'm wondering if it can really be as simple as "do what step one says, do what step two says."

If there is anybody out there who wants to take a look at this and give me an opinion as to how difficult this will be to install and use, I'd appreciate it.
Toronto Spider is offline  
Sponsored Links
Advertisement
 
Status: T-5 Days to Unicorn Paradise
AO Platinum Member
 
TheGrimJeeper's Avatar
 
Join Date: Sep 2009
Location: France
County: Riviera
Posts: 38,319
It looks pretty straight forward TS.

I guess IBM are looking to expand the learning of Watson by offering this for free.

IBM Watson
TheGrimJeeper is offline  
(Post Link) post #3 of 21 Old 03-05-16 Thread Starter
Status: -
AO Gold Member
 
Join Date: Nov 2002
Location: Canada
Posts: 6,209

Member car:

1973 Spider 2000

Well that's what I thought, but then I second-guessed myself figuring that maybe I was missing something complex and codey. Like I can see myself suddenly at at command prompt wiping out my entire OS by accident. (Yes, I know that's highly, highly unlikely, but you get the gist.)
Toronto Spider is offline  
(Post Link) post #4 of 21 Old 03-05-16 Thread Starter
Status: -
AO Gold Member
 
Join Date: Nov 2002
Location: Canada
Posts: 6,209

Member car:

1973 Spider 2000

Looking over the instructions again, I seem okay until I get to step 5.
Quote:
Install project dependencies and build browser application:

$ npm install && npm run build
That sounds an awfully lot like coding. Or am I missing something?
Toronto Spider is offline  
Status: T-5 Days to Unicorn Paradise
AO Platinum Member
 
TheGrimJeeper's Avatar
 
Join Date: Sep 2009
Location: France
County: Riviera
Posts: 38,319
npm is similar to a make, I guess....?

So compiling, installing and such like?
TheGrimJeeper is offline  
(Post Link) post #6 of 21 Old 04-05-16 Thread Starter
Status: -
AO Gold Member
 
Join Date: Nov 2002
Location: Canada
Posts: 6,209

Member car:

1973 Spider 2000

I guess.

I know I'm out of my depth here. Maybe I'll just have to bite the bullet and try it. I really could use some good transcription software that doesn't require training. Dragon is utterly useless. From the test I ran with Watson, it seems better. It's far from perfect and some of the mistakes are hilarious, but even if it's 60 or 70 per cent accurate it would be really helpful.
Toronto Spider is offline  
Status: Sticking up for MGs everywhere!
AO Silver Member
 
David WB's Avatar
 
Join Date: Apr 2014
Location: United Kingdom
County: Lancaster
Posts: 3,769

Member car:

147 & GQV

"npm" is just the manager for JavaScript packages. The command is just the installation line, followed buy a build command to be able to run the program.

All of the complicated stuff after it is merely opening the websocket so the commands are sent back and forth to the server, and to tell "cf push" what to actually *do* with the program (in the yml document in the current directory). You definitely won't blow up the system with those commands.
David WB is offline  
Status: Gormless [email protected]*
AO Gold Member
 
Paddy OPlastic's Avatar
 
Join Date: Feb 2014
Location: Ireland
County: Cork
Posts: 8,465

Member car:

Alfa GT JTD 2004

Quote:
Originally Posted by Toronto Spider View Post
I guess.

I know I'm out of my depth here. Maybe I'll just have to bite the bullet and try it. I really could use some good transcription software that doesn't require training. Dragon is utterly useless. From the test I ran with Watson, it seems better. It's far from perfect and some of the mistakes are hilarious, but even if it's 60 or 70 per cent accurate it would be really helpful.
Have you tried Google's speech-to-text in the Chrome browser or on Android?
Paddy OPlastic is offline  
(Post Link) post #9 of 21 Old 04-05-16 Thread Starter
Status: -
AO Gold Member
 
Join Date: Nov 2002
Location: Canada
Posts: 6,209

Member car:

1973 Spider 2000

Quote:
Originally Posted by David WB View Post
"npm" is just the manager for JavaScript packages. The command is just the installation line, followed buy a build command to be able to run the program.

All of the complicated stuff after it is merely opening the websocket so the commands are sent back and forth to the server, and to tell "cf push" what to actually *do* with the program (in the yml document in the current directory). You definitely won't blow up the system with those commands.
Thanks. That's really helpful.

As I said, I was only joking about blowing things up. My main concern was that I really don't know what I'm doing, and I wasn't sure it was just typing a command line or if there was something much more intricate that was required of me.

Quote:
Originally Posted by Paddy OPlastic View Post
Have you tried Google's speech-to-text in the Chrome browser or on Android?
Not yet. I don't use Chrome or Android. I generally try to keep Google out of my life as much as possible (outside of the search engine functionality which I haven't really been able to replace. DuckDuckGo only works so well and hasn't proven itself a completely reliable substitute).

I liked the idea that I could run Watson locally. I'm sure it's still going to be reporting back to Big Blue, but for whatever reason, that doesn't bother me as much as telling on me to Google.
Toronto Spider is offline  
Status: Gormless [email protected]*
AO Gold Member
 
Paddy OPlastic's Avatar
 
Join Date: Feb 2014
Location: Ireland
County: Cork
Posts: 8,465

Member car:

Alfa GT JTD 2004

Quote:
Originally Posted by Toronto Spider View Post
I liked the idea that I could run Watson locally. I'm sure it's still going to be reporting back to Big Blue, but for whatever reason, that doesn't bother me as much as telling on me to Google.
You won't be running Watson any more locally than Google unless you've got a pretty powerful rack full of computing goodness and a large bank account for the license.
Paddy OPlastic is offline  
Status: Sticking up for MGs everywhere!
AO Silver Member
 
David WB's Avatar
 
Join Date: Apr 2014
Location: United Kingdom
County: Lancaster
Posts: 3,769

Member car:

147 & GQV

Quote:
Originally Posted by Paddy OPlastic View Post
You won't be running Watson any more locally than Google unless you've got a pretty powerful rack full of computing goodness and a large bank account for the license.
Even worse is that Watson may eventually be running you if it starts getting all clever and shirty.

I understand though, IBM seems more 'respectable' than many other tech companies. I'm sure there are those that would object and say it is just as dirty as any other company, but the general perception is that they don't play 'big brother' and share details about you willy nilly. They come across much more like researchers who are interested in research and pushing boundaries, rather than drilling into your life for maximum profit (despite it being a for-profit company).
David WB is offline  
Status: Gormless [email protected]*
AO Gold Member
 
Paddy OPlastic's Avatar
 
Join Date: Feb 2014
Location: Ireland
County: Cork
Posts: 8,465

Member car:

Alfa GT JTD 2004

Quote:
Originally Posted by David WB View Post
Even worse is that Watson may eventually be running you if it starts getting all clever and shirty.

I understand though, IBM seems more 'respectable' than many other tech companies. I'm sure there are those that would object and say it is just as dirty as any other company, but the general perception is that they don't play 'big brother' and share details about you willy nilly. They come across much more like researchers who are interested in research and pushing boundaries, rather than drilling into your life for maximum profit (despite it being a for-profit company).
It's true that IBM hasn't traditionally been interested in collecting vast amounts of data like Google and Facebook. However Watson needs to digest data to become more clever so you can be sure that IBM isn't opening up access altruistically. Anything you send it will become part of it.
Paddy OPlastic is offline  
(Post Link) post #13 of 21 Old 04-05-16 Thread Starter
Status: -
AO Gold Member
 
Join Date: Nov 2002
Location: Canada
Posts: 6,209

Member car:

1973 Spider 2000

Quote:
Originally Posted by Paddy OPlastic View Post
You won't be running Watson any more locally than Google unless you've got a pretty powerful rack full of computing goodness and a large bank account for the license.
I didn't get that impression from the instruction page, although I'll admit I didn't explore the accompanying documentation thoroughly. It says it's released under an Apache 2.0 licence, so I would suspect that negates the big bank account concern.

As for privacy concerns, I realize that overall it's a lost cause. Everything is reporting back to the mothership these days. (Windows 10 I'd love to upgrade to keep current, but I'm not handing over total control to MS. Yes from what I've read, I realize 8.1 is also spying on me, but at least I can control when/if I install updates.)

It's funny how my perception of Google has shifted. For a while years ago they were offering free long-distance services in Canada, as part of their speech recognition research. I ended up using that service even though I hated the idea of being recorded and turned into data. But while it was in the research stage, it didn't bother me too much, especially if I was circumspect about which calls I used it for. But these days now that the company is all about selling ads and linking online identities and profiles it makes me squeamish.

I realize IBM will be using everything I would potentially run through it as research fodder and will attempt to use the knowledge it gains to make future technologies and profits, but since it's at the fuzzy, nebulous stage I'm more willing to go along with it.

Ideally I'd prefer to actually buy a good solution that works, but at this point, it doesn't seem to be available. As I mentioned above, I actually tried Dragon, and for my purposes it doesn't work.

Last edited by Toronto Spider; 04-05-16 at 15:46.
Toronto Spider is offline  
(Post Link) post #14 of 21 Old 04-05-16 Thread Starter
Status: -
AO Gold Member
 
Join Date: Nov 2002
Location: Canada
Posts: 6,209

Member car:

1973 Spider 2000

Quote:
Originally Posted by Paddy OPlastic View Post
Have you tried Google's speech-to-text in the Chrome browser or on Android?
Just out of curiosity how well does it work? I may be willing to forego my privacy concerns if the tool actually performs as I require. (Yes, I can be a hypocrite if the need is truly there. )
Toronto Spider is offline  
Status: Gormless [email protected]*
AO Gold Member
 
Paddy OPlastic's Avatar
 
Join Date: Feb 2014
Location: Ireland
County: Cork
Posts: 8,465

Member car:

Alfa GT JTD 2004

Quote:
Originally Posted by Toronto Spider View Post
I didn't get that impression from the instruction page, although I'll admit I didn't explore the accompanying documentation thoroughly. It says it's released under an Apache 2.0 licence, so I would suspect that negates the big bank account concern.
The only code released under the Apache license is the code which records your voice, sends it to IBM and receives back the transcription. All the heavy processing happens within an instance of Watson which is running in an IBM data centre somewhere.
Paddy OPlastic is offline  
Status: Gormless [email protected]*
AO Gold Member
 
Paddy OPlastic's Avatar
 
Join Date: Feb 2014
Location: Ireland
County: Cork
Posts: 8,465

Member car:

Alfa GT JTD 2004

Quote:
Originally Posted by Toronto Spider View Post
Just out of curiosity how well does it work? I may be willing to forego my privacy concerns if the tool actually performs as I require. (Yes, I can be a hypocrite if the need is truly there. )
It's scarily good.
Paddy OPlastic is offline  
(Post Link) post #17 of 21 Old 04-05-16 Thread Starter
Status: -
AO Gold Member
 
Join Date: Nov 2002
Location: Canada
Posts: 6,209

Member car:

1973 Spider 2000

Quote:
Originally Posted by Paddy OPlastic View Post
The only code released under the Apache license is the code which records your voice, sends it to IBM and receives back the transcription. All the heavy processing happens within an instance of Watson which is running in an IBM data centre somewhere.
Ah, that makes a lot of sense.

Quote:
Originally Posted by Paddy OPlastic View Post
It's scarily good.
I may just have to try it. Good to know I've got options.
Toronto Spider is offline  
Status: T-5 Days to Unicorn Paradise
AO Platinum Member
 
TheGrimJeeper's Avatar
 
Join Date: Sep 2009
Location: France
County: Riviera
Posts: 38,319
As Po'P says, Androids speech-to text is very usable, it's just not
as convenient to dictate sms/emails as it is type them in a lot of
circumstances. So after initially trying to switch to it, I slid back to
stabbing at the touch keyboard.
TheGrimJeeper is offline  
(Post Link) post #19 of 21 Old 05-05-16 Thread Starter
Status: -
AO Gold Member
 
Join Date: Nov 2002
Location: Canada
Posts: 6,209

Member car:

1973 Spider 2000

Does it allow me to input .WAV files?

I have an ongoing need to transcribe audio files lasting anywhere from 15 to 60 minutes all from different speakers. I'm the world's worst typist.

As I said, I ran file through the Watson test and the results weren't terrible. Some things were wildly off, but in general I could make sense of what was outputted. In comparison, Dragon produced things that might be classified as modern poetry run back and forth through a language translation program. Actually, Dragon's results were pretty hilarious. A colleague and I ended up laughing our heads off at the results. Admittedly, Dragon wasn't created with my needs in mind. It's supposed to work for a single person who has time to train it, but sadly that's not my situation.
Toronto Spider is offline  
Status: T-5 Days to Unicorn Paradise
AO Platinum Member
 
TheGrimJeeper's Avatar
 
Join Date: Sep 2009
Location: France
County: Riviera
Posts: 38,319
Watson sound like a better fit TS, but in
general speech-to-text with cloud computing
behind it works very well these days.
TheGrimJeeper is offline  
Status: She's happy to pootle
AO Platinum Member
 
Join Date: Jan 2001
Location: United Kingdom
County: Surrey
Posts: 16,723
Tor - you have a PM re: Dragon ;-)
Scudetto is offline  
Reply

Go Back   Alfa Romeo Forum > Misc Lounges > Community Discussions > The Technology Section

Tags
speech , text , watson

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page



Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off

 
For the best viewing experience please update your browser to Google Chrome