On Fri, May 29, 2009 at 4:48 PM, David K Watson
<[email protected]>wrote:
> It's a funny quote, but it ignores the fact that Wolfram Alpha
> was never intended from the start to be a "Google killer".
> From a PC World article:
>
>> The first thing Wolfram Research co-founder Theodore Grey wants you to
>> know is what Alpha is not: It is no "Google killer," as it's been called by
>> some reports. In fact, Alpha is very, very different from a search engine.
>
>
"Some reports" is an understatement. Every single report, article,
write-up, commentary, and everything else I've seen on WolframAlpha, except
from Wolfram Research itself, compares it to Google. I think 30 seconds
with WA would convince someone that there is no comparison between the
two... but I think WA doesn't really protest too hard because its so
difficult for them to explain what they are and why they've included some of
the data they have.
> Alpha is fairly good at what it is supposed to do, which
> is to present data systematically and to do analysis of
> that data that you are unlikely to find on any web page.
>
I keep having mixed feelings about this. I keep wanting to feel this is
true, but every single query I've tried that I thought was relatively
simple, WA has totally failed me. WA doesn't really understand the
relationships of much of its data, while there are many many webpages out
there that do.
To give just a simplistic example. Ask WA about Cassablanca, and say you're
talking about the movie. It knows that Humphrey Bogart was in the movie.
Ask it about Humphrey Bogart... and it knows that he was an actor, but
doesn't tell you that he was in Cassablanca. Contrairwise, go ask IMDB
about Humphrey Bogart, and it will tell you that he was in Cassablanca. Ask
Google about Humphrey Bogart, and it refers you to IMDB which will have that
kind of relationship.
I'm pleased, however, that WA has been improving the queries that work.
When it first launched, I tried "distance earth mars", which worked fine,
and "distance venus mars", which happily told me the distance from Earth to
both Venus and Mars. Last week when I tried this, it gave me the expected
results.
> And it ties in really well with Mathematica if you want
> to do a more sophisticated analysis of that data.
>
I would hope that it does! {: That said, I can't comment on this, since I
haven't used Mathematica over a decade, and even then I was probably doing
things far far simpler than what WA provides.
It does explain part of what dissapoints me about WA. The results are
very... flat. I'll do a stock comparison, look at the graphs, see an
outlier... and get frustrated that clicking on the outlier doesn't work. I
can only assume that Mathematica would provide that kind of integration.
> There is plenty of room for Alpha to coexist alongside
> Google, and I expect that eventually Alpha results will
> turn up in Google search results for some queries in
> much the same way Wikipedia results do now.
Actually, I expect that Google will increasingly be doing similar work on
its own, trying to do its own semantic parsing. While the "expert reviewed
data" that WA has is good, what the web is increasingly trying to say is
that "very interested non-experts" are providing nearly as good data, and
that there are tools out there that are connecting this data together.
I keep wanting to like WA... but I just can't find any results from it that
I care about.
*************************************************************************
** List info, subscription management, list rules, archives, privacy **
** policy, calmness, a member map, and more at http://www.cguys.org/ **
*************************************************************************
