Thursday, September 27, 2007

Intel Mash Maker shows promise, but isn't there yet

I've spent the last couple of days playing around with Intel's Mash Maker, trying to see if it is a tool I would feel comfortable putting in front of a business masher. As with so many other mahsup tools, the answer is "No." Not only is it much too fragile at the moment (Their server has been down more than it has been up over the past couple of days.) but it still requires users to know HTML and XPath to scrape content.

However, unlike Google's Mashup Editor and Yahoo! Pipes, I can see the skeleton of a real business-user tool in the making.

For those of you who may not have read about Mash Maker, it is a semantic web-based HTML screen scraper developed at Intel's Berkley Research Lab. The idea behind the tool is to leverage the power of everyone in the world to classify, tag and otherwise bring under control the chaos of content that is now the web. Then when the content is tagged and understood, allow said content to be mashed with other content while providing hints about other available mashups.

Here is an example. There is nothing new about plotting all the good burrito restaurants near a specific location. We've been mashing up those sorts of sites for a couple of years now. However, what if you wanted to find the good burrito restaurants that were within walking distance of a tire repair shop? If you had to use Google or Yahoo! to develop this mashup, it wouldn't be worth your time. However, with Mash Maker you can create the mashup in minutes.

Run a search for burrito joints in your area code. Show the results as a table (A 'standard' mashup provided by Mash Maker) and copy them. Run a search for tire repair shops in your area. Paste in the burrito restaurant information and map the results. Now you have plotted on one map the burrito joints and the tire repair shops. It's faster to build the mashup than it would be to run two Google Map searches and tally the results on paper.

Why do I say the tool isn't yet ready for prime-time? Server crashing problems aside, getting to the point where Mash Maker understands the content in a web page is difficult. Any page that Mash Maker doesn't understand, and that includes nearly every website that I visited, needs to have a data extractor written to scrape out the content. Creating a data extractor requires the user to know how to write XPath expressions, and how to understand the relationship between the page's HTML and the rendered content. Sorry, but that's too much to ask of a business user. No, not because the business user is stupid, but because these users are experts in their subject area, not in HTML and XPath.

I have to conclude that "Mashups for the masses" is still only a pipe dream. However, I think Intel is on the right track. While Mash Maker isn't ready for my business mashers yet, it could be in the future.

No comments: