17 June 2009

Roundup of new advanced search tools

There's been several recent feted releases of unique search tools. Though none are sea changes, they add interesting possibilities and complement existing methods (usually your search-engine-of-choice + Wikipedia).

First up, Wolfram|Alpha: structured answers to a hand tooled domain of data. The best use for this is when Wikipedia's search fails, usually when you have a combination of words within the domain you're interested in. Sometimes a Google search into Wikipedia can resolve that (e.g. search for site:wikipedia.org rival "Franz Liszt" from Google). The biggest benefit of W|A is the clearly formatted results hyperlinked to deeper, related searches.

Second, the entertainingly naive Google Squared. It's still in Google's "Labs" area, so it's in alpha or earlier and gives results that you'd expect more from someone's Google mash-up or Greasemonkey script. Results are structured like W|A but culled from data scraped from the web rather than hand-picked. I hope it gets more attention from the developers because dynamic and emergent knowledge is more scalable than edited knowledge. When new information appears on the web, GS doesn't need to have updates added like encyclopedia yearbooks. W|A does.

Finally, of lesser note, TextRunner [ via Slashdot ] from the University of Washington allows very simple natural language queries and, like Google Squared, finds answers in an unedited corpus of web pages so is more easily scalable (if a bit slow). Like most basic NL search engines, queries are in the form WH-WORD VERB NOUN. While Google et al. have won the day by de-emphasizing semantic knowledge, it's good to see that tagged information extraction is still being researched.

[ posted by sstrader on 17 June 2009 at 10:39:38 AM in Internet , Science & Technology ]