Tonight I've been playing with voice recognition.... And it's actually working

[TheAnimus]

Mostly.

It's been years. I've used Dragon Simply Speaking back in the 90s, I've tried all sorts, and been disappointed wholly.

Tonight I've been playing with the developer preview of Windows Phone 8.1, that pretty much anyone can put on their phone, with the beta personal assistant cortana.

What has amazed me, is how it is able to work so well, offline.

"Turn on Flight Mode"
...
"Play Clean Bandit A and E"

The latter command is impressive, because the music track is titled A+E. It has deciphered that is what I want. Very impressive, considering this is done offline, on a tiny mobile phone processor.

It's really super impressive compared to my experiances with Siri, which given T-Mobile's data speed in the City has been subpar.

But what's even more fun is the fact it has really good understanding of certain actions, I told it:

"Remind me when I get to the office to book my bicycle in for a service"

It decided this should mean that when I arrive at a location it knows is my office (geofencing), it reminded me.... That's pretty cool.

NLP I think is needed to really take the next jump into usefulness for interaction of any kind of complexity of event. It will be interesting to see how well it works with "tell me if traffic is bad".

[TooNice]

I wonder how well voice recognition can handle different accent nowadays. No English speaking human has an issue understanding me, yet no one has been able to identify my accent (when speaking English) which is probably due to the fact that I kept bouncing between various countries during my formation years. Dragon Simply and the likes definitely did not like it when I tried back in the 90s, and I more or less resigned to the fact that, unless a program can learn a completely new accent, that I will have to live without voice recognition. But if I could, I've love to have it on a DSLR. I have ran into a couple of instance when I want to make some setting changes which requires going deep into a menu when I have already framed the shot, and wished I could just order it instead of having to spend time through the menu settings.

[Spud1]

It's fun and rather cool, and nice to see this sort of tech evolving.

I still think it's of *very* limited used in the real world, mostly due to "Voice recognition anxiety" or "not wanting to talk to my phone in public", but when home alone or in the car, it can be amazing. I use the Google Now version to set reminders all the time, as its quicker/easier than typing it in. That is about it really, as I don't find it useful for anything else..but like you, I am amazed/pleased with the results and what it can do.

MS seem to have taken the best bits of siri/google now and merged them together, and despite its daft name (really MS needs to sort out its marketing/branding team) all that i've seen of it has worked really well.

I hope they integrate it into the XBO at some point to get past the NLP issues it has - i'm fed up of say "xbox off" and then remembering i need to say "Xbox Turn Off" before it will do anything.

[shaithis]

My brother has been using the various versions of dragon over the years due to disability. About 3 years ago he felt that they had cracked it. It does take a bit of training to get it used to you but the success rate once trained is damn good.

[Dooms]

I had pretty much the same experience as you TheAnimus, back in early 2000 having to setup Dragon for people at work and having a play with it myself. Having to read the entire of Alice in Wonderland for it to even start deciding to know what I'm on about.

Fast forward 10 years and just using basic commands on Google Now or using free software with no training and it getting pretty much spot on. Things are getting better but I think I'll stick to typing for the time being

[TheAnimus]

MS seem to have taken the best bits of siri/google now and merged them together, and despite its daft name (really MS needs to sort out its marketing/branding team) all that i've seen of it has worked really well.

Well in this case it was originally the code name, Cortana, the personal assistant to you Master Chief in Halo.

But as this leaked out, people really like the name, and frankly, I think it's good. Google Now is not just nondescipt, but hard to actually remember or think of as a 'personality thing'. Having a name like Siri or Cortana, is clever, the words are very recognisable, think of it as a header for a packet with a very high hamming distance (I'm thinking your the spud that did electronic engineering and signals right? If not my analogy might be useless lol).

So you address the thing that way. It knows immediately that it is an action. XBOX is another great one, however, people will also refer to it "Oh I'm just playing with the xbox", no one will say "I'm just using the Cortana".

This is why I think it's a great choice, it's got the branding tie in with arguably one of MS's most well known entertainment successes. I'm amazed MS didn't mess it up, I was expecting something like Bing Personal Windows Phone Assistant.... Live!. Compared to Google, I think they did better in branding. However they will be behind Apple as they have spent so much money with their meet Siri stuff.

It's also in my experience a lot better than Google Now, which wouldn't handle "Remind me when I get to the office, to hassle Ed about the outstanding...".

I've also been playing with the API, which sadly I've gone and broken my code (and doubt I'll have personal project time for a while to fix). But before I messed up my VCD file, I was able to say:

"Cortana, did I turn the lights off?"
"Alex, you left the lights on in the Living Room and the Study, would you like me to turn them off?"
"Yes please"
"You have no lights left on, would you like me to switch them on when you get home"
"Yes please"
"When you arrive at Home, I will switch the Living Room lights to 1500k and the Study lights to 2500k"

I've still got a lot to do adding the proper chains of events, when you can cross from one scenario to another (state machine time!) but it's very well designed API, and something that is lacking in the others I find. The downside is that there is quite a long pause between commands, effectively she finds that the action is supported by my application, which then has to be spun up (I've not implemented fast resume!), my application then has to talk to a website, which then talks to a little hub device on my home network, this is of course not very fast. As a result there is a long 3 second pause between my command a response.

[Moby-Dick]

Very impressive As long as it returns HAL style error codes.

"I'm sorry Alex , I can't let you do that"

[TheAnimus]

I do like that if you ask her to sing you a song, she sings daisy daisy.

[Spud1]

(I'm thinking your the spud that did electronic engineering and signals right? If not my analogy might be useless lol).

Close, I did go to the same uni as Agent who I *think* did that course (but then again I may have confused his course ), but I do get the analogy all the same.

I think Cortana is just a bit cumbersome thats all, whereas "Siri" may be a daft name but its quite slick and simple, rather than a more complicated "Cortana". I take your point though, and not being a Halo fan I never got the Halo reference I'm personally not that keen on the whole personification thing either but I totally get why that works (or is even necessary) from a marketing/sales perspective.

Love the sound of what you are doing with the API though, and thats the sort of area where voice control can /really/ start to take off..in the home. Home automation is finally starting to go mass market with Hive and recently "Tado" (http://www.tado.com/gb/, which i've just ordered) and it wouldn't be that much work to hook something like that up to a voice recognition system.

Geofencing+cortana+Tado+ZWave..could be awesome. I'll be surprised if the Xbox Two doesn't incorporate something similar

[TheAnimus]

At the end of the day its just one extra phonym, which I think is worth it for the marketing tie in thing.

Did you see my post about LILA, a London kickstarter, which is about to re-launch, they messed up a bit, as kick starter doesn't let you sell more than 10 "units", their idea was bluetooth low power door sensors, simple, not exciting, but I had a pledge for a kit of 10, plus a WiFi bridge, for £90. Suddenly the automation options become better. These things need to be simple, mass market, and open enough that anyone can integrate them anyway they wish.

I took a look at Hive and frankly, thought it was a pants as Nest. £200 to let me remotely turn a relay off or on, in the words of hot fuzz *jog on*. I'm not giving you all this information and control on live energy usage (this is very valuable information, see how certain things like electricity are priced) and not even have a documented available API.

What I want is to be able to control each room. I've looked at a few kits, but at the moment they are all quite large and clumbersome. I don't have thermostatic valves at the moment, so feel as if I am just burning money, I figure that I might as well go electronic, as some of them don't require me to drain the water. I've been looking at these in particular: http://www.zwave-store.co.uk/eurotro...tor-thermostat

[mikerr]

despite its daft name (really MS needs to sort out its marketing/branding team)

Cortana, the personal assistant to you Master Chief in Halo.


Google Now is not just nondescipt, but hard to actually remember or think of as a 'personality thing'. Having a name like Siri or Cortana, is clever, the words are very recognisable, .

Never knew the Halo reference for the cortana name...

Siri is such a better name than Cortana (which I keep forgetting),
but even that is much better than saying "ok google" or "hi galaxy" on samsung phones.

[virtuo]

Sure my grandad used to drive a Ford Cortana

[Frankenfurter]

On the one hand it's great to see this technology improve, but on the other hand voice control cocking up is the best part of voice control.

I'm genuinely torn on this issue.

[Zak33]

Siri is such a better name than Cortana (which I keep forgetting)

that'll be due to marketing... if the name Cortana hasd been Apple's, you'd remember it!