Is human powered search the next big thing?

I’d be well within reason to suggest that it would be somewhat difficult to use the web without a good search. Google has been so successful that their name has become a verb in the English language that defines using their service to search for something on the web. Traditionally, search engines compile a ginormous index of all the data their crawlers can collect on the web and then provide the fastest possible way for you to compare a search phrase with the content in their index. As good as the services have become, they haven’t been able to truly differentiate between relevant and irrelevant results for a large number of search phrases. You still have to sift through a few pages of results a lot of the time.

Lately, I’ve come across a few search engines that are attempting to take a different approach to the content relevancy problem. Instead of a proprietary algorithm that attempts for calculate relevancy, e.g., Google’s PageRank, they are letting relevancy be determined by human review. I decided to give RankNoodle.com a try and see if I could get better results with something powered by a bunch of other people than I can by simply typing in the search phrase in the search box of my browser and instantly getting a bunch of hits.

Here are a few review criteria I decided to test out:

1. Does it work?

   First things first. Can I find information with their service? I decided to try and search for a few things I was in need of while I was writing this post. I didn’t want to just type in random search queries and browse the results. I wanted to try and find specific things that I need. For instance, I’ve been wanting a WordPress plugin that will tell me the word count of my blog posts. I searched for “wordpress word count plugin” and was pleasantly surprised to find a good option as the 2nd result. There are not 1000s of results like when you search on Google or Yahoo. I’m not exactly sure if users contribute the results or if RankNoodle uses crawlers and then just lets the users vote the results up or down. Either way though, the results page contained perfectly sufficient resources and I found what I wanted right away.

   RankNoodle.com still has work to do on basic functionality though. All the search results I obtained, for all my different search queries, advertise as having 41 comments. Nothing happens when I clicked on comment though so I’m not sure what is going on there. For a search engine however, I’m not terribly concerned about what other people say about the URLs. I’d rather just see what I want in the title and description listed with the URL and be able to make a quick decision from that about which URL to click on next. For the record, nothing happened when I clicked on “Share” or “Bury” either. I don’t know if that is a browser problem (FireFox 3.0 on Mac OS X), or just that they still have work to do on the interface.

2. Is it fast enough?

   I think to for any search engine to be competitive, human powered or not, it has to be fast. There simply isn’t time to wait for results to fetch and display when other options can do the same thing in fractions of a second.

   I noticed that common search items are cached on RankNoodle.com so I thought I’d search for something that isn’t likely to have many users searching for it. I spent a lot of time this week working with wxPython and XRC resources (Yes, I’ll probably be needing to write a post about this too.) I wanted to know how to extend a wxPython Widget and have it automatically be loaded with the XRC handlers. I searched for “wxpython xrc custom class“. As I expected, the search did take a little extra time. As it stands now, it’s a little too slow to be competitive with other search options but I imagine if the concept is something that users will like and use, that performance can be tackled as the service grows.

3. Usability

   I think overall, RankNoodle.com adheres to what a typical user would expect of any search engine. There is a search box on every page and results are not obscured by other page features. It would be nice if your current search terms were still listed in the search box on the results page though. It’s difficult to remember exactly what you typed and retype all but a portion when refining search results. I do like how image, video and related articles are listed on the side. In my opinion, that is a lot more worthy usage of the extra white space than filling it up with ads.

   That brings me to another subject. Search result pages are considered “articles” by RankNoodle.com. In addition to your search results, you get a general description and comments/questions by users for that article. I found it funny that the wxPython search I did when I first started testing had a question that was answered with “This is not a good place to ask this question.” Perhaps certain articles will get better results.

Anyway, overall, I think social or human powered searching has a lot of potential. For me to spend more time with it, it will have to be fast, integrated into my browser, and get me to the results I want. I can’t think of a better way to go up against current major search engine players though than by taking a completely different approach.