I just came across a pretty neat site and service for creating vertical search engines of your choosing. Called a ‘swicki’ the service and capabilitiy is provided by Eurekster, a company founded about two years ago around the idea of personalized and social search. The ‘swicki’ implementation was first released in November 2005.
classification information science knowledge management
metadata ontologies OWL RDF semantic web web2.0 XML [ ?]
NOTE: As you conduct searches using the form above, you will be taken from my blog to http://swisher-swicki.eurekster.com. To return, simply use your browser back button.
What in Bloody Hell is a Swicki?
According to the company:
Swickis are a new kind of search engine or search results aggregator. Swickis allow you to build specific searches tailored to your interests and that of your community and get constantly updated results from your web or blog page. Swickis scan all the data indexed in Yahoo Search, plus all additional sources you specify, and present the results in a dynamically updated, easy to use format that you can publish on your site – or use at swicki.com. We also collect and organize information about all public swickis in our Directory. Whether you have built a swicki or not, you can come to the swicki directory and find swicki search engines that interest you.
Swickis are like wikis in that they are collaborative. Not only does your swicki use Eurekster technology to weight searches based on the behavior of those who come to your site, in the future, your community – if you allow them – can actively collaborate to modify and focus the results of the search engine. . . . Every click refines the swicki’s search strings, creating a responsive, dynamic result that’s both customized and highly relevant.
A 10 Minute Set-up
I first studied the set-up procedure and then gathered some information before I began my own swicki. Overall the process was pretty straightforward and took me about 10 minutes. You begin the process on the Eurekster swicki home page.
- Step 1: You begin by customizing how you want the swicki to look — wide or narrow, long or short, and font sizes and a choice of about twenty background and font color combinations. I thought these customization options were generally the most useful ones and the implementation pretty slick
- Step 2: You "train" your search (actually, just specify useful domains and URLs and excluded ones). Importantly, you give the site some keywords or phrases to qualify final results accepted for the site. One nice feature is to add or not blog content or the content of your existing web site
- Step 3: You then provide a short description for the site and assign it to existing subject categories. Code is generated at this last step that is simple to insert into your Web site or blog, with some further explanations for different blog environments.
You are then ready to post the site and make it available to collaborative feedback and refinement. You can also choose to include ads on the site or look to other means to monetize it should it become popular.
If a public site, your swicki is then listed on the Eurekster directory; as of this posting, there were about 2,100 listed swickis (more in a next post on that).
For business or larger site complexes, there are also paid versions building from this core functionality.
SWISHER: Giving it My Own Test Drive
I have been working in the background for some time on an organized subject portal and directory for this blog called SWISHer — for Semantic Web, Interoperability, Standards and HTML. (Much more is to be provided on this project at a later time.) Since it is intended to be an expert’s repository of all relevant Web documents, the SWISHer acronym is apparent.
One of the things that you can do with the Eurekster swicki is run a direct head-to-head comparison of results with Google. That caused me to think that it would also be interesting when I release my own SWISHer site to compare it with the swicki and with Google. Thus, the subject of my test swicki was clear.
Since I know the semantic Web reference space pretty well, I chose about 75 key starting URLs to use as the starting "training" set for the swicki.
This first version of SWISHer as a swicki site, with its now-embedded generated code, is thus what appears above. In use it indicates links to about 400,000 results, though the search function is pretty weak and it is difficult to use some of my standard tricks to ascertain the actual number of documents in the available index.
To see the swicki site in action, either go to http://swisher-swicki.eurekster.com, click on the SWISHer title, or enter your search in the form above and click search.
Now installed, I’m taking these capabilities for a longer road trip. The test drive was fun; let’s see how it handles over rough terrain and covering real distances. I’ll post impressions in a day or so.