The much-hyped search engine – sorry, “computational knowledge engine” – Wolfram Alpha launched over the weekend. Its use of databases and semantic search should be particularly exciting for journalists because a) it searches parts of the ‘hidden web’ that most search engines don’t reach (i.e. databases); and b) it has the potential to throw up quick answers to questions about relationships and facts that Google is also not great at.
Now, note that I say “potential to” – Wolfram Alpha is in its very early days. Below I talk about what it can do now; I expect it will be years before it truly becomes a game-changer.
The ideal search engine for a pub quiz
Wolfram Alpha deals in facts. It deals with natural language questions quite well, but only if you’re looking for something very specific about something well known.
‘How old is Barack Obama?‘ worked, for example; as did the same enquiry for Gordon Brown, and indeed, possible future Labour leader Alan Johnson. It could also give me the ages of Madonna, rapper Chuck D and David Beckham, but not Newcastle midfielder Kevin Nolan.
How about ‘5 largest countries in the world?’ Not immediately, but it did suggest ‘largest countries’ which brought up rankings by area, population and GDP .
It works well for buildings, providing answers to the questions ‘When was Notre Dame built?’ and the rather less well known ‘When was the Rotunda built’ (the Rotunda is a building in Birmingham); and it can give you answers about works of literature.
Demographically it can not only tell you the populations of countries and cities (and detect that you’re in the UK and so not interested in, say, Birmingham, Alabama), but plenty of extra detail, along with tourism information and trade data. You are blessed if you live in the US, where you can instantly find out the average wage of a journalist or nurse (it didn’t recognise ‘midwife’), unemployment rates and other detail – but for the moment if you live in any other country it appears you will have to look elsewhere.
Other features that may be useful include the ability to find out what the weather was like in a particular place on a particular date or how much a certain amount of money was worth at some point in the past; while poets and scrabble junkies might appreciate the linguistic functions that allow you to search for anagrams, synonyms, definitions, and words that sound like other words. You can also use it to complete aphorisms and nursery rhymes.
From a journalistic perspective, these features are a time-saver if you don’t fancy browsing through almanacs and biographies for the same facts. But that’s it. And it’s not clear where the information is coming from or how accurate it is (Karen Blakeman, whose review is worth reading, told me it gets some things wrong, “even chemical structures”) – that’s the advantage of Google or Wikipedia: you can evaluate the credibility of the source relatively intuitively; Wolfram, however, presents itself as the source, and where links are given in ‘Source Information’ these are often just to homepages).
It’s a calculator as much as a search engine
Beyond facts, Wolfram is very good at calculations. Searching for ‘convert 500 euros to pounds sterling’, for example, works nicely – but then it also does on Google; it’s just most people don’t know that.
It can calculate fuel use and fuel costs; it can even calculate nutritional values or calories of a given combination of foods, including brand names. You can imagine some immediately useful applications for these if you are covering, say, motoring, transport, health or food – but more generally, any publisher with a website featuring that sort of information would do well to start looking at Wolfram’s API to see how their website might automatically generate that information.
It also goes without saying that the money and finance functionality is enormously useful for anyone involved in financial journalism.
And it knows that 2+2=4.
A search engine waiting for the web to catch up
Sadly, if you want to ask more interesting questions that connect some of those worlds you will struggle (although Jeff Mummert has some cute examples).
‘UK education budget 2007′ – while understood – results in no answer. Change that to ‘UK education expenditure‘ and you do get a graph showing the trends up to 2004 – but no data more recent than that. ‘UK Politicians born after 1960′ or ‘politicians related to JF Kennedy’ sadly don’t work. And while it knows what year Dickens was born, it can’t turn that around to tell you what writers were born in that year.
But that’s because Wolfram is an engine waiting for the world to catch up. The technology is enormously impressive – really, game-changingly important. But the material it has to work with is, currently, sparse. As one reviewer commented, it is:
“like a roomful of idiot savants. Each knows a scary amount about a topic. And, unlike a such a roomful, WA also knows how to recombine and compute what each of the savants knows. But if the room doesn’t have the savant you’re looking for, you get back nothing but a “Huh?”
To add to that, if the savants speak a different language, you can’t get them to talk.
In 5 or 10 years the structure of data may have changed enough – and enough people will have volunteered – to make this very powerful indeed; and there is also a part of me that hopes the very existence of Wolfram – and Google’s semantic-friendly updates in response - along with many other mashups, will increase the cultural pressure on governments and corporations to make data available in a machine-readable format (witness the UK Conservatives publishing expenses in real-time).
Then we can ask “How much has my MP spent on dog food?“
If you’ve tried Wolfram and found some other uses, or can see further potential, please let me know.