Google Beats Wolfram|Alpha On Its Own Turf
I was told about a strange and intriguing fact: that Tuscany, Italy is at the same latitude as Ontario, Canada. Based on my incomplete understanding of European geography (as well as that of the Great Lakes,) this seemed kind of improbable. Given that Wolfram|Alpha had just debuted, this looked like exactly the kind of query the “computational engine” had been designed to handle. Unfortunately, it was not able to provide any response for either location.
Query: latitude tuscany, italy
Response: Wolfram|Alpha isn’t sure what to do with your input.
Query: latitude ontario, canada
Response: Wolfram|Alpha isn’t sure what to do with your input.
Now, this is a just-introduced beta product, so to show that the engine is capable of producing latitudes:
Query: latitude ontario
Response: Assuming Ontario (California, USA) result: 34.0528
Ontario, Canada was not one of the choices available in the “more” drop-down menu.
So, next I turned to Google, a general purpose search engine that returns spidered web results and is not optimized for these kinds of computational queries.
Query: latitude tuscany, italy
Response excerpt (top result):
These provide fast and efficient connections with the rest of Italy. Tuscany’s smaller towns are… Tuscany Co-ordinates: Latitude: 43.44: Longitude: 11.25 …
Query: latitude ontario, canada
Response excerpt (top result):
Look-up Latitude and Longitude – Canada. This table gives the latitude and … 45° 22′ N 63° 16′ W Yarmouth AP 43° 50′ N 66° 5′ W ONTARIO Belleville 44° 9′ …
While Google’s results might be considered a little messy compared to Wolfram|Alpha, they did actually return a result in each case, and that result contained the reasonably correct latitudes for Tuscany and Ontario. At a glance, I was able to confirm that the original statement had merit. Some part of Tuscany, Italy really is at the same latitude as some part of Ontario, Canada. (I’m glad I wasn’t wagering money on this!)
Wolfram|Alpha is new, and understandably incomplete. The Semantic Web is the future, but the experience we can hope for is tied to not just accuracy, but also flexibility in how we are allowed to interpret the data. Google works because it can draw on all of the data on the web, no matter how poorly organized, or whether it has been formally curated. The downside of this approach is that you cannot reliably organize the results of a query performed on such data. Hence, Google’s generic ranked list of snippets for any search. Wolfram|Alpha offers amazingly customized results, but relies entirely on known semantically correct data, which must be curated and edited. Unfortunately, that approach will not scale very well for many kinds of information due to the level of intervention required.
A lot of people are saying that to get the most out of Wolfram|Alpha requires that you don’t try to use it like Google, or expect the same results. And we are being told not to even think of comparing this new technology to Google on any level. In this case, the query was one where the computational engine should excel. And eventually I think (or hope) it will. But even Google’s clumsy and disorganized results provided a meaningful answer, and I think there’s a real lesson in that.
Nothing excessively fragile has ever survived on the web, no matter how cool it is. The Semantic Web is coming, but the winners will have to be robust.
