On Thu, Jan 20, 2011 at 09:27, Dylan McCall
<dylanmccall@xxxxxxxxx> wrote:
> On Mon, Jan 3, 2011 at 5:04 PM, Mark Shuttleworth <
mark@xxxxxxxxxx> wrote:
>> MPT is worth talking to, as he will have good thoughts about how we can
>> get a rich base of textual data for the search. Anything in the Dash
>> applies pretty much directly to Ubuntu Software Centre too :-) so it's
>> worth chatting to MPT and the developers of USC.
http://menukeywords.appspot.com/
That's the quick prototype for the little web app / survey thing I was
thinking of earlier. (I just shared it with MPT, too, in an email, so
he might have some thoughts). Pretty simple start which could be grown
in a few directions. Under the hood it has a database with a bunch of
applications I had installed on my system at the time. (And, of
course, it stores all the words people think of to describe those
applications).
We can manipulate the database to extract all kinds of information. In
particular, I think it would be interesting to see how certain
keywords distribute over different categories. It could also help to
identify where problems lie with the idea. (Are the categories
descriptive enough? Do they meet peoples' immediate expectations? Are
some apps misusing them, or not using them when they could be?)
In the final thing, the set of applications should probably be a lot
smaller; just stuff installed by default or featured in Software
Centre.
As it is, I estimate I could trick about ten people to enter a few
keywords for about three applications. (That's the point where you get
a cake recipe and run off to the nearest microwave in excitement).
There are probably lots of ways this could work better and be more
interesting (maybe to turn that ten into twenty), so if anyone has
thoughts or suggestions I'd be thrilled.
This work goes very much into the direction of shaping a semantic desktop here, i'm in awe and i can't help feeling extraordinarily happy about what you are doing here!
To overcome the burden of scientificalizationalistic terminology, i'd like to introduce a not-so-new paradigm, which could guide the efforts here towards a more natural interaction:
Natural Language Processing.
One step into the direction of NLP would be to allow not only keywords but also phrases. When communicating with each other, we usually don't limit our speech to single words, we speak in sentences.
For printing for example, i would love to tell the computer "stop printing!"
Until today, computers wouldn't listen, if you spoke to them in such a fashion, neither when addressing them via voice, nor via keyboard input.
You would have to enter a formal aka formatted instruction, before anything would actually happen.
Now, with "semantic" command interfaces that are already capable of understanding and autocompleting keywords, this is coming closer, and again i can't help it but throw in my 2ct about how computers could become intelligent, if we only sat down and made it happen.
I can't imagine a friendlier interface than one which speaks my language, so all i'd have to know in order to use it is my language.
Having a collaborative "teaching" interface like the one you are presenting here would of course be a vital part in shaping such an experience, but i dare not overwhelm anybody again with my science fiction..
