I've seen lots of reviews of Wolfram|Alpha now - I've only recently had a chance to play around with it, and it's certainly different. Let's start with the blindingly obvious - it's not a search engine. Yes, I know that it's called an engine, though they say that it's a computational knowledge engine, and this is the key difference. Too many people are trying to use it in the way that they use Google et al, and this just doesn't work - it's like expecting an encyclopedia to give you the address of the local takeaway; doesn't mean that there's anything wrong with the reference tool - it's just being used in the wrong way. It's like asking a librarian what pi is to 150 places - the librarian can find a resource that gives you the answer, while a mathematician can work it out for you. Both entirely valid ways of answering the question, but totally different answers. So, before we can progress with an understanding of WA it's absolutely key to put aside thoughts of 'Google killers'. (If you want a 'killer' reference, I'd be more inclined to look towards it being a Wikipedia killer.)
Most reviewers (including me) will start a search by looking for themselves. We normally know where we're referenced, how many references there are and so on. However, if you do this over at WA you get information about the names 'phil' and 'bradley'. We learn that Phil and Bradley are male given names, latest information for US births (1976 for phil and 2007 for bradley), history of US births from 1880 to date, estimates for current US population and estimated current age distribution. The source data for this comes from 6 different places, all of which are listed. The age of some of the data is disappointing - 1976 isn't exactly that helpful. It's also a shame that we're only getting content from the United States - adding in UK or England just confuses the engine, and we get no results.
Trying a date also gives really different information (remember the computational element here!), because if I type in a date such as April 9 1865 I learn a lot about the date - that it fell on a Sunday, it was 144 years, 1 month and 10 days ago. (Lovely little cut and paste function there which I only discovered when trying it!). It's also the 99th day of the year, the date that the American Civil War ended, sunrise/sunset information and the phase of the moon. A search on the same date in Google emphasises the historical information, rather than anything else.
Another example of the computational element - a search for population england china india france germany provides a summary of the total population for those countries, highest and lowest, tabular data, population history, and a huge slew of data on the demographics. This content is taken from many different (listed) sources, and they're fairly current. However, while it did give the population for my local town, this was dated to 2004. On the other hand, the same search in Google gives me a more recent, but much less precise figure.
A search for 25 australian dollars sterling in WA gave me an answer of £12.49, an exchange history and additional conversions. Google didn't give me a proper result until I put 'in sterling' into the search. Google gave me a figure of £12.51 and it was difficult to get more content from Google that was comparable to WA.
I liked WA, but think it's been insanely overhyped. I'm concerned about the sources used, the US bias and the date of the information returned to me, but hopefully this will improve over time. The bottom line is that if you need an answer, ask Google. If you need an answer worked out, ask WA.
I think if the database behind WA were made open in the same way wikipedia is it would stand a much better chance. Not sure if that would be possible or not, but it would be interesting to know.
Posted by: Library Web | May 20, 2009 at 05:18 PM