I don’t even mind the blank text box as a starting point–but they should be teaching users how to make structured queries so that their frustration from misinterpreted NLP queries goes down as they use the system.

In any case, I don’t think that casual users need a computational engine.

In fact, judging by the examples in their gallery, I’m not even sure Wolfram Alpha is aiming at casual users. Here’s what the preview site shows as “a few things to try”:

* enter any date (e.g. a birth date)
june 23, 1988
* enter any town (e.g. a home town)
new york
* enter any two stocks
IBM Apple
* enter any calculation
$250 + 15%
* enter any math formula
x^2 sin(x)

That doesn’t strike me as resonating with casual users.

I’m sure they could spend a lot of time designing a precise interface that it turns out nobody likes or feels is intuitive. I am personally a proponent of the “blank text box” type systems, as seen in Google, Ubiquity, Spotlight, etc. The problem WA has is that they have no idea how people will use it, because 1) they have little to no data, and 2) a “computational engine” is a new paradigm.

Of course, this brings the question of whether the users should conform to the query language (power users, high learning curve) or the query language should conform to the user (casual users, low learning curve), which is ultimately driven by their monetization strategy–and who knows what that will be!

The reasons it upstaged WA: it released publicly while WA was still in private beta, and it offered access to much of the same curated data as WA. It doesn’t support any computation of inferences, but most of the WA examples don’t do that either.

I think WA could do a lot more–but only if they can get past the interface issues. NLP is not the right interface for a computational engine!

Good to have, but hardly an upstaging of WA, don’t you think?

Or am I not getting something?