Show HN: Star Trek-like computer with voice interface | Hacker News Reader

Show HN: Star Trek-like computer with voice interface

github.com

98 pointsrob_mccann38 comments

I think there will be some cool applications with this and a Rasberry PI. I'm floating around this half-baked idea of plugging Rasberry PI's into the light sockets of my home (since they're all central to the room [better for audio and microphone quality] and negates the need of running wire all over the place) so that I can have voice controlled computers in all my rooms and play music/make calls/automate my home/etc. Being able to ask the computer some random question and get an answer without having to go to my computer would be insanely cool to me :)

That's the idea, this is literally just the first concept there's a whole bunch of cool things to be done. I'd love to make it possible for others to build their own extensions too.

Awesome project, Rob! It's amazing what can be done with some cheap hardware and some ingenuity + existing services over the internet. We just blogged about your project on Hack Things. http://www.hackthings.com/star-trek-like-computer-with-raspb...

Agreed with this sentiment. Something that could ease its way into the home improvement segment is an embedded linux device with microphone input, speaker output, powered off 120VAC, fits in a wall box, and input for more microphones to form an array for better listening.

That's the idea behind getting it to run on the Raspberry Pi. Having it choose the mic closest to the person that says 'computer' would be cool.

Agreed here. Guess I'm thinking the best way to get this functionality into homes though is with a board spin of a embedded linux platform (i.e. RPi) that people could install into a wallbox.

A bluetooth speaker-like package would probably be fine too.

Android 4.1, and above, has offline voice recognition. I'm not sure about the current state of Android support for Raspberry PI, but you wouldn't have to reinvent anything.

I've had exactly the same thoughts myself. I was thinking about using TP-Link WR703n's instead of Pi's but if they get the model A out in quantity soon enough, it would be a better choice.

Powerful computers that are so cheap you can permanently purpose them to tuning on and off a single light-bulb are a bit of a game changer if you think about it.

I'm guessing the next raft of patent trolling will follow the form "household item but with a computer/wifi in it".

I've already been working on something very similar for my car as I was sick of taking my eyes off the road to change the music etc. I already have a working prototype, but there's still some work to be done before its good enough to publish.

Hmmm, on Mac, might be possible to simplify using system's dictation (have no idea if they have API for that, probably no, but maybe there is a way to hack it) and system's voice. So then it might be even faster, and would require mostly just Wolfram|Alpha API key.

Sounds like a good idea to me, I wonder if Windows has a similar API too?

Yes it does SAPI[1]. There's a few for linux but nothing that's become any kind of standard unfortunately, festival tends to have the best voices but is a bit of a pain to use inside another program in my experience.

[1] http://msdn.microsoft.com/en-us/library/ee125663%28v=vs.85%2...

It originally used festival but it was a pain to install on the Pi (which I want this to run on) and the voices I tried weren't as easy to understand. I'd love for it to be an option, though.

Since 1995: http://en.wikipedia.org/wiki/Microsoft_Speech_API

The voice is terrible, though.

In OSX 10.8 you can just hold Cmd (or is it option?) for a few seconds to get dictation for any text input system-wide.

For TTS the say command has all voices available.

Double tap 'fn'.

OH. So that's what that sound was from. I've accidentally triggered it and couldn't figure out how that happened.

This has now been added

The future is an API mashup. Neat. It does make me wince a bit thinking that the actual techniques driving all of this are locked up tight, and subject to going away whenever the owning business decides to do so. Hopefully the actual future is a little more open.

Agreed, it'd be cool to develop some kind of verbal querying for the semantic web rather than relying on a proprietary knowledge base - but that's a whole other project for another day.

There is a view of the future where we query Google (or Siri) by voice and it just gives us exactly what we want.

I think a better "vision" is an open voice input engine that then connects, or even bridges, other systems via APIs.

I particularly like the idea of a service such as Zapier bridging a infinite universe of APIs to connect things together -- "computer request Uber black car when I receive a Skype call from Jim."

An input engine could create a marketplace of APIs. Commercial services, such as Uber, could profitably offer their APIs for free. Other services without a specific product may run their APIs on a cost per use basis. A mass market cost-per-use API could charge a few pennies per inquiry but still be very profitable.

Perhaps what I am describing could represent the demise of the search engine as we know it.

Neat. I'm working on a system and API for question type (factoid, polar, description, etc.) prediction and polar question answering (polar questions are 'yes-or-no' questions; Wolfram Alpha doesn't attempt these). Let me know if you are interested.

Is it on github?

I would like to hear more about your project.

It is Siri minus a lot of the language processing. I'm glad to see interest in the topic since this is the type of thing I am currently working on a http://www.stremor.com

But Wolfram alpha has a lot of things missing, and lots of the things it does are way longer than I want to hear a computer read to me.

Maybe if the author combined this with our TLDR api, and our Sentence parsring API, and our Fact search API...

(yes if you use all of our API's you can build this in about and hour and have it do a lot more) hint hint

Feel free to fork and improve, I'd be interested to see how it would work with your APIs.

You're claiming to do better than wolfram alpha? Are there any online demos of your technology? The better search through semantics has been thrown around for years, but so far it's never been found to beat the traditional approach. POS and grammatical comprehension has all the usual brittleness. The best or the best is CoreNLP, but that's GPL so unlikely to be in your product unless you've licensed it.

Check us out. We don't use CoreNLP or NLTK. we have our own because everything else sucked.

And yes I do better than wolfram alpha.

Why don't you have pricing, dev docs... or... anything other than marketing speak on your website? I see zero actionable/useful information there.

Edit: And I looked around for like five minutes. I assumed you would have some easy-to-start sort of thing there since you're telling small Github projects to use you.

We will be in the azure data market monday or tuesday.

As CTO I am always the last to know that we weren't already there.

maybe if you install their TLDR plugin - it will help you find out just the right info you need on their website ;-)

more seriously though - I felt the same. Lots of marketing talk, but little actionable stuff to see this put to practice. The Samuru search engine didn't seem to respond with anything vastly better than google or duckduckgo, but I didn't really compare searches, or can't think of an intelligent way to do so.

Is there an extensible framework for processing Wolfram Alpha-like natural language queries (presented as text) locally on Linux?

python's nltk, I've not yet looked into it but do plan to.

You should totally hook Pi-Voice up to NinjaBlocks.com and all your connected devices!

1. do you work for ninja blocks? 2.Just looked at Ninjablocks do they have microphones?

I don't work for NB, but I think you could probably get an add-in board/USB solution for audio input. It's based on a Beagle Bone (ARM/Linux). It might even accept Arduino shields?

So it's open-source Siri in Python? Cool!

Reply on news.ycombinator.com