John Timmer brings up my two biggest complaints about Wolfram|Alpha. The first is that it’s even harder to identify the source of information than it is in Wikipedia, the other is what happens when searches fail:
A bad Web search typically brings up results that help you refine your search terms; a bad Alpha search returns nothing, and it’s not clear that there’s an easy way to fix that.
Here’s a simple way: have Alpha fall back on library data. One example he offers, “global bioethanol production,” is perfect for both library reference and bibliographic collections.
Google Book Search offers a few promising hits for that query, as does Google Scholar. And if we imagine that search in the context of faceted collections, we’d be able to identify subjects the search phrase is associated with and offer additional access points. Add to that some crowdsourcing opportunities for users to expand the knowledge base by identifying an item or piece of text that answers their question and we’ll have a real competitor to Google.