As promised, in this second post I’m going to tell you about an unusual feature of Noflail Search. Suppose you remember having made a yummy smoothie recipe that you saw in cookwithnikki.wordpress.com a few months ago, and you want to find the recipe again online. Go to noflail.com, put it in Advanced mode, enter the query “banana mango peach smoothie”, choose Bing as the search engine to run the query on, and enter cookwithnikki.wordpress.com into the Site box. The top of the window should look like this:
Now click Search. At the time of this writing, the query produces no results. (Things may have changed by the time you try it, of course.) But Noflail Search doesn’t just say that there are no results. It also provides a zero-result analysis, in a box that appears at the bottom of the Search History panel and looks like this:
The third entry in the analysis says that there are no posts in the CookWithNikki blog containing the word mango and the word peach. That tells you that you remembered the ingredients wrong: there was either mango or peach, but not both. The first entry in the analysis tells you that there are 11 posts with banana, mango and smoothie, and the second entry that there are 2 posts with banana, peach and smoothie. You may remember that it was a tropical-sounding recipe, so that mango is more likely than peach, and you may then click on the first entry. That will run the query “banana mango smoothie” and show its 11 results in the Results panel. The smoothie you were looking could be the Banana Mango Flaxseed Smoothie recipe of the first result. (Again, things may have changed by the time you try this, and your first result may be different.)
I don’t think there is currently any search engine that provides zero-result analysis. Please correct me if I’m wrong. But the concept is actually not new, it has been known in academia for decades. What’s new is the parallel algorithm that we are using to compute the analysis by issuing queries to a Web API across the Internet, in only a few seconds. Previous implementations were sequential and would have taken minutes rather than seconds in the same setting.
What we are really proud of, though, is that we have taken a useful concept buried in academia for a very long time and made it available to everybody in Noflail Search.
Interesting