1787: "Voice Commands"

This forum is for the individual discussion thread that goes with each new comic.

Moderators: Moderators General, Prelates, Magistrates

rmsgrey
Posts: 3655
Joined: Wed Nov 16, 2011 6:35 pm UTC

Re: 1787: "Voice Commands"

Postby rmsgrey » Thu Jan 19, 2017 11:21 pm UTC

(because someone has to say it)

It's all Greek to me...

DaphneDiane
Posts: 4
Joined: Fri Feb 01, 2013 7:13 am UTC

Re: 1787: "Voice Commands"

Postby DaphneDiane » Thu Jan 19, 2017 11:56 pm UTC

Reminds me that I used to aliases set up so that would automatically rewrite the X11 keyboard map to switch to Dvorak, including h.soav, ih, om, &c. Also had an alias ,.pyf that switched me back to Qwerty.

ej159
Posts: 2
Joined: Mon Jun 16, 2008 6:49 pm UTC

Re: 1787: "Voice Commands"

Postby ej159 » Fri Jan 20, 2017 1:19 am UTC

This comic got me thinking: are there languages that are easier to design voice recognition systems for?

In these days of deep learning it probably makes very little difference but does anyone know?

User avatar
PinkShinyRose
Posts: 835
Joined: Mon Nov 05, 2012 6:54 pm UTC
Location: the Netherlands

Re: 1787: "Voice Commands"

Postby PinkShinyRose » Fri Jan 20, 2017 6:44 am UTC

ej159 wrote:This comic got me thinking: are there languages that are easier to design voice recognition systems for?

In these days of deep learning it probably makes very little difference but does anyone know?

Any language with a native phonetic writing systems with very few loanwords? So most languages that aren't English or Chinese?
ps.02 wrote:
somitomi wrote:Hmmm, is it weird if I can read cyrillic letters despite not speaking any language that uses it? I mean I know why I can read Greek letters (thanks, engineering school), but I had to actually bother learning cyrillic ones.

Well, and to be fair, once you pick up the Greek alphabet, you're already halfway to sounding out Cyrillic.

And if you also pick up the Latin script, you're almost there already.

Mikeski
Posts: 1113
Joined: Sun Jan 13, 2008 7:24 am UTC
Location: Minnesota, USA

Re: 1787: "Voice Commands"

Postby Mikeski » Fri Jan 20, 2017 8:02 am UTC

PinkShinyRose wrote:
ej159 wrote:This comic got me thinking: are there languages that are easier to design voice recognition systems for?

In these days of deep learning it probably makes very little difference but does anyone know?

Any language with a native phonetic writing systems with very few loanwords? So most languages that aren't English or Chinese?

There are a lot of things that would make languages easier or harder for computers to interpret.

Large number of phonemes? Similar phonemes? Multiple fricatives? Harder. (One of the easier tests for audio over-compression is "can I still tell an 'F' sound from a 'V' sound"... and a lot of voice recognition is done in that same FFT'ed frequency domain.)

I would guess that tonal languages would be harder than non-tonal ones.

A large number of homophones would require more knowledge of grammar, rather than being a simple dictionary match.

Varied accents could make things tough; interpreting various English speakers (British, American, Caribbean, Indian, etc.) could be a nightmare. Most humans can't do it reliably.

Yu_p
Posts: 51
Joined: Wed Oct 12, 2011 9:00 am UTC

Re: 1787: "Voice Commands"

Postby Yu_p » Fri Jan 20, 2017 10:05 am UTC

Heimhenge wrote:
ChurchSkiz wrote:I just realized how I'm going to be "an old guy" to my grandkids.

"Grandpa what is that rackety thing on your desk with all those letters?? You mean you actually have to touch the letters to make words???? Why not just let the voice module take over?"

"I prefer the old clicking of the keys!"


You have GOT to be kidding me. I mean, I know there are/were some famous writers who stuck with typewriters after the dawn of the word processor, but I never understood the motivation.

I struggled through writing a student tech manual the old way. Then going back and drawing in the equations/subscripts/superscripts etc. with a Koh-i-Noor ink pen (not to mention going through a lot of correcting fluid). Took frikkin' forever. If I'd have waited a few years I could have done it with a word processor in probably half the time. When I got my first word processor, a DOS-based app called New York Edit that came on a single floppy disc, my life changed in a "quantum leap" kinda way. Well, at least my writing life ... didn't make me any less a geek. But geez ... a choice of fonts, different sizes, inflection marks. Writing was actually fun. Best of all ... no ribbon tool bars!


Different use-cases. For publication / summaries I write use LyX or LaTeX, but I still use paper as everyday Work-Log, and for doing derivations. I burnt myself bad trying to do everything the digital way during my diploma-thesis (lack of quick sketches, too easy to break a meaningful ordering of the entries, suprisingly low overall usefulness of text-search, etc.).

Digital writing is quite superior, when you already know the details of what you will write, and reorganizing/formatting becomes a primary concern. However, I can perfectly relate to the appeal of more analog methods as part of the workflow, especially when – as with authors working under a publisher – you aren't even supposed to do the parts, where the digital tools can shine. Mostly it is about "focusing on the content, not on the tool" for me.

Heimhenge wrote:Writing was actually fun.

When I want writing to be fun, I use a calligraphy pen :D

User avatar
squall_line
Posts: 170
Joined: Fri Mar 20, 2009 2:36 am UTC

Re: 1787: "Voice Commands"

Postby squall_line » Fri Jan 20, 2017 8:06 pm UTC

Mikeski wrote:
PinkShinyRose wrote:
ej159 wrote:This comic got me thinking: are there languages that are easier to design voice recognition systems for?

In these days of deep learning it probably makes very little difference but does anyone know?

Any language with a native phonetic writing systems with very few loanwords? So most languages that aren't English or Chinese?

There are a lot of things that would make languages easier or harder for computers to interpret.

Large number of phonemes? Similar phonemes? Multiple fricatives? Harder. (One of the easier tests for audio over-compression is "can I still tell an 'F' sound from a 'V' sound"... and a lot of voice recognition is done in that same FFT'ed frequency domain.)

I would guess that tonal languages would be harder than non-tonal ones.

A large number of homophones would require more knowledge of grammar, rather than being a simple dictionary match.

Varied accents could make things tough; interpreting various English speakers (British, American, Caribbean, Indian, etc.) could be a nightmare. Most humans can't do it reliably.


Especially Scottish accents

See also: https://www.youtube.com/watch?v=5FFRoYhTJQQ

Which is a nice circular path back to Mikeski's movie quote (thank you for that, by the way; it was in my head and you kindly posted it for me. :))

Sly
Posts: 3
Joined: Fri Jan 20, 2017 8:13 pm UTC

Re: 1787: "Voice Commands"

Postby Sly » Fri Jan 20, 2017 8:21 pm UTC

As a Dvorak user, I'm laughing so much at this one. Even more so when I tried decoding it (the wrong way,) and got this:
ö.ak fööfrh yhpj a vhnv
Notice something odd?
(Yep: German Keyboard.)

User avatar
somitomi
Posts: 756
Joined: Fri Nov 06, 2015 11:21 pm UTC
Location: can be found in Hungary
Contact:

Re: 1787: "Voice Commands"

Postby somitomi » Sat Jan 21, 2017 10:24 am UTC

Sly wrote:As a Dvorak user, I'm laughing so much at this one. Even more so when I tried decoding it (the wrong way,) and got this:
ö.ak fööfrh yhpj a vhnv
Notice something odd?
(Yep: German Keyboard.)

I wonder what I'd get by mapping Dvorak over the Hungarian layout.
Image
―◯‐◯ FG Discord ◯‐◯―

User avatar
Soupspoon
You have done something you shouldn't. Or are about to.
Posts: 4060
Joined: Thu Jan 28, 2016 7:00 pm UTC
Location: 53-1

Re: 1787: "Voice Commands"

Postby Soupspoon » Sat Jan 21, 2017 12:46 pm UTC

niauropsaka wrote:Τηατ'σ νοτ ηος τηισ ςορκσ! Τηατ'σ νοτ ηος ανυ οφ τηισ ςορκσ!

I'm so glad I can read that. Algebra is useful for something, kids! Learn the basics, and you can probably muddle through the rest!

(Also addendumed my original message, now I have power again. Yes, I did get ninjaed. ;) )

User avatar
jc
Posts: 356
Joined: Fri May 04, 2007 5:48 pm UTC
Location: Waltham, Massachusetts, USA, Earth, Solar System, Milky Way Galaxy
Contact:

Re: 1787: "Voice Commands"

Postby jc » Sat Jan 21, 2017 6:23 pm UTC

rmsgrey wrote:(because someone has to say it)
It's all Greek to me...

And it's all Chinese to the Greeks. ;-)

Anyway, I've been disappointed that nobody has tried to divert the discussion to the question of whether the claims for the Dvorak layout have any statistical validity. I've read a number of comments from people who tried to track down the history of the claims, who've said they never uncovered anything but tiny samples and/or "studies" whose details are proprietary and unavailable to "outsiders". Which isn't to say that the claims are wrong, of course; it's just saying that we might want to be a bit skeptical about claims that originate from people who stand to make a profit from selling things to people who believe the claims. But I guess it's not what you'd call a major issue of the day.

rmsgrey
Posts: 3655
Joined: Wed Nov 16, 2011 6:35 pm UTC

Re: 1787: "Voice Commands"

Postby rmsgrey » Sun Jan 22, 2017 1:43 am UTC

jc wrote:Anyway, I've been disappointed that nobody has tried to divert the discussion to the question of whether the claims for the Dvorak layout have any statistical validity. I've read a number of comments from people who tried to track down the history of the claims, who've said they never uncovered anything but tiny samples and/or "studies" whose details are proprietary and unavailable to "outsiders". Which isn't to say that the claims are wrong, of course; it's just saying that we might want to be a bit skeptical about claims that originate from people who stand to make a profit from selling things to people who believe the claims. But I guess it's not what you'd call a major issue of the day.


Yeah, not many people are going to base their choice of keyboard layout on which is most efficient - rather they're going to base it on what they have ready access to, or what they get provided at work (granted, you can probably come to some arrangement whereby you get a special keyboard, but that's going to require extra effort to arrange) - or what allows them to feel superior to other people...

User avatar
chridd
Has a vermicelli title
Posts: 846
Joined: Tue Aug 19, 2008 10:07 am UTC
Location: ...Earth, I guess?
Contact:

Re: 1787: "Voice Commands"

Postby chridd » Sun Jan 22, 2017 1:49 am UTC

jc wrote:Anyway, I've been disappointed that nobody has tried to divert the discussion to the question of whether the claims for the Dvorak layout have any statistical validity. I've read a number of comments from people who tried to track down the history of the claims, who've said they never uncovered anything but tiny samples and/or "studies" whose details are proprietary and unavailable to "outsiders". Which isn't to say that the claims are wrong, of course; it's just saying that we might want to be a bit skeptical about claims that originate from people who stand to make a profit from selling things to people who believe the claims. But I guess it's not what you'd call a major issue of the day.
I mean, since QWERTY was designed to slow down speakers so that early voice recognition software could have a chance to keep up, Dvorak is clearly superior, and studies aren't needed. :)

rmsgrey wrote:Yeah, not many people are going to base their choice of keyboard layout on which is most efficient - rather they're going to base it on what they have ready access to, or what they get provided at work (granted, you can probably come to some arrangement whereby you get a special keyboard, but that's going to require extra effort to arrange) - or what allows them to feel superior to other people...
...or be different from other people, or fit in with other people, or feel more comfortable (subjectively), or what they learned back when they were younger, or...
Of course, "ready access" is probably true for most QWERTY users.
~ chri d. d. /tʃɹɪ.di.di/ (Phonotactics, schmphonotactics) · she · Forum game scores
mittfh wrote:I wish this post was very quotable...

umagp
Posts: 2
Joined: Mon Apr 11, 2011 4:51 pm UTC

Re: 1787: "Voice Commands"

Postby umagp » Wed Jan 25, 2017 12:59 am UTC

When I type this in, I get 2 errors "oxay google send a tekt". My translation would be: "sbat ussupd ;dlh a kbxk". Am I missing something? Is this an Azerty Dvorak? Or some improbable Colmak Dvorak layout?

Edit: Almost six years between first and second post. Takes Dvorak to bring me out of my hidey hole.

User avatar
Soupspoon
You have done something you shouldn't. Or are about to.
Posts: 4060
Joined: Thu Jan 28, 2016 7:00 pm UTC
Location: 53-1

Re: 1787: "Voice Commands"

Postby Soupspoon » Sat Jan 28, 2017 7:37 am UTC

umagp wrote:Edit: Almost six years between first and second post. Takes Dvorak to bring me out of my hidey hole.
If you would not yet have made a second post, otherwise, does this represent proof that Dvorak increases the speed of posting? ;)

(no answer needed, this side of 2022, if you want to maintain your reputation... :mrgreen:)


Return to “Individual XKCD Comic Threads”

Who is online

Users browsing this forum: No registered users and 96 guests