The platform will remain open and become entirely free for everyone.
A nice model, this. Facebook gets a huge corpus of semantic data from the users of the app, and everyone else gets an ever-more-sophisticated classifier for their commands.
That's a win/win for everyone. Well done and congrats!
Maybe I'm cynical, but talk is cheap. I remember when Facebook bought face.com. From the face.com blogpost announcing it:
> "we love you guys, and the plan is to continue to support our developer community.”
...the API was shut down some time later. I would dispute the claim the developer community was supported as a result of that deal. A blog post saying the platform will remain open isn't a commitment, it's a vague, hand wavy promise.
The reason I don't think that will happen here is Wit's pricing model - it's completely gratis to use, with the proviso that your commands and associations are public. You cough up cash if want that to remain private.
As long as new users are using Wit, Facebook continues to get value from the product. Closing it down wouldn't gain them much.
- If you have a paying subscription, you won’t be charged again and your instance(s) will remain private.
- Your open instance(s) will remain open
- If you have open instance(s) that you want to make private (for free), please wait for a few days until we roll out this feature.
I'm a paying customer and had plans to pay them significantly more (I am building an app that will be released in a few months that will heavily rely on it). Personally I'd rather see it stay a paid service since I want an SLA and support.
That said, Wit.ai is a great service, the team is friendly and responsive, and I congratulate them.
"The reason I don't think that will happen here is Google Reader's pricing model - it's completely gratis to use, with the proviso that your commands and associations are public... As long as new users are using Reader, Google continues to get value from the product. Closing it down wouldn't gain them much."
Sarcasm aside, companies sometimes put such an emphasis on "focus" that commitments to users of their non-core products end up falling by the wayside.
Not to mention Moves who promised your data would remain separate and then promptly changed the TOS to allow Facebook access to your data. I believe that was the last app update too.
A friend of mine's reaction, quoting bits of the linked post:
<Mc> 18 months ago, we started Wit.ai with the vision that no solution is to
be found in a closed, centralized, managed approach.
<Mc> That is where today’s big news comes in: Wit.ai is joining Facebook!
I wasn't aware of Wit.ai until today's news... but now after checking out their web site and API, it will be probably a great addition of their API to Parse.com!
Potentially some killer features can come up by combining both systems... Siri on steroid with cross platform solution! ouch!
Pure speculation here, but it doesn't seem crazy to think that this is in part for use by the Oculus team. While voice commands are a nice to have on current mobile phones - they are likely to be a much bigger deal for head mounted displays.
This acquisition actually makes a lot more sense to me than Facebook buying Oculus. Oculus' ostensible business model was selling hardware to users, which is totally orthogonal to Facebook's business model of selling users to advertisers. IMO there's absolutely no way Facebook bought Oculus just to get the revenue from selling Oculus hardware, but I still don't see what their plan would be for integrating that hardware into their existing business model. I can definitely see how a natural-language API would fit in.
Unlike vomit inducing 'VR' (I own both dev kits) and over broad ideas of 'IoT' - speech and intelligent assistants (ie, siri/cortana/ok google, alexa, jibo, etc) is the future of computing.
If the Oculus team wanted to acquire Wit.ai, they could've acquired Wit themselves (as they recently did with Nimble). Keep in mind Oculus is an independently operating subsidiary within Facebook.
Wit.ai is a great service and it is great news Facebook will be keeping them open. Would be a great fit for some of their other dev services like Parse. I built an app that took some users voice commands and pokes fun at the users crudely and it took so little time to build because of Wit.ai.
Congrats to the Wit.ai team! For once I'm glad it was facebook who acquired them. I've been truly happy with Parse and can only begin to wonder how facebook will integrate all the developer tools they are acquiring.
I'm assuming that puts an end to my dreams of Wit.ai supporting device-local voice recognition with context-limited vocabularies to drive user interfaces though...
I found pocketsphinx pretty easy to work with. I use it for command voice recognition for home automation stuff. It's even pretty accurate when using limited vocabulary models, after some tweaking. There are python bindings, though they lag behind the C api slightly, and even the C api is well-commented and the code is clean.
Do you have a write-up/blog post or article about pocketsphinx? How do you create/train the vocabulary models? If you use Text-to-speech too (CMU Flite, eSpeak), maybe you know some good resources there as well?
I haven't written anything up yet. To summarize, I use the standard acoustic model (hub4wsj_sc_8k) with a combination of keyword activation and a fixed grammar (in jsgf format). It's normally listening for a wakeup keyword, and when it find one, switches into grammar mode until it hears a complete utterance, or times out, then switches back to keyword mode. It works pretty well, though tuning the keyword sensitivity is annoying.
The pocketsphinx-specific code is actually quite simple:
You can see the keyword and jsgf files in that directory, for reference. The pronunciation dictionary is generated from one of the standard dictionaries, selecting just the works present in the grammar.
Note: If you look in the whole directory, there's a mix of decent code and ugly hacks in there, and I didn't make any attempt at making it customizable. It's just for me. Btw, it all runs on a raspberry pi.
Basically, you can define vocabulary sets to check against, and you provide a callback to know when a word has been detected. Depending on the complexity of the vocabularies, you might need to spring for the Rejecto plugin to improve the results.
I was never a user, but when I checked them out, did think their offering showed great promise. I suppose the offer from Facebook must've been too good to say no.
Hadn't heard of wit.ai before. Seems very nice. One huge thing missing from the site is any information about supported languages. Still don't know....
Always loved their interface. I'm interested to see if Facebook will make it easy for the tech to be shared with their Oculus developers as well to create VR agents or controls that are easy to interact with via voice.
Was just telling a non-tech friend about Wit.
Try, explaining NLP and classifiers to someone who isn't quite sure what dictation is, made for a quick ride up the chair lift at Kirkwood.
Congrats to Wit, though I can't help but be a little saddened that the self-hosted/embedded licensing may be less likely to roll out as a result. I've been playing with the platform and could definitely find some use, but most of our target market is unlikely to accept cloud based services, especially those associated with data-mining organizations. Fingers crossed that such licenses/installations are still on the roadmap.
I don't think that's a fair label in this case since they are keeping the product. For context, the blog at http://ourincrediblejourney.tumblr.com highlights the contradiction when an acquisition announcement is filled with flowery language about the adventure of building the product and other excitement while simultaneously shuttering the product.
Obviously everyone has the right to shut down their products, but it'd be nice if they said "We are sorry that we are going to be unable to continue offering this service".
It's well known that anyone a bit into SV and its surroundings is aware of this "meme" and founders are often placing bets whether they will include it in their acquisition-of-product-PR or not.
I can imagine that this is how it feels when you go from a few people in a apartment/basement/house to a company with an office and eventually acquisition. (I've only been through two of the three so far) Regardless of how cliche it might sound at this point it's understandable they always use the same term.
A nice model, this. Facebook gets a huge corpus of semantic data from the users of the app, and everyone else gets an ever-more-sophisticated classifier for their commands.
That's a win/win for everyone. Well done and congrats!