Friday, November 20, 2009

Search Queries Unpacked

I'm at the Illinois Educational Technology Conference in Springfield, IL waiting for my presentation videos to upload.  Going over the wireless network is excruciatingly slow, so I thought I'd check my blog stats.

One of the free services I use with this blog is sitemeter. One of the most interesting features of this service is "referrals:" the url where readers were just before coming to this blog. I haven't tallied the numbers, but it appears that about half of those drawn to this blog arrive as a result of a Google search.

For example, here's the url that brought one of the last readers to the Internet Search Challenge blog:
If you unpack that url (or click it), you can tell what query the person used to find the blog: internet searches for students. You can see the blog listed in the search results for that query. Currently it's #8 although that can change as time goes on.

That's a pretty good query, even though the stop word "for" is unnecessary. Here's a larger sampling of recent queries:

Detecting biases in writing
challenges searching by keywords
detecting bias activity
query two common interpretations internet
challenges in searching with keywords
optimal number of search terms
how has getting info easier with internet
"they formed an angel band" lyrics

Several observations occur to me in unpacking these queries.

First, even though queries are not formatted particularly well, they still led to relevant information (I'm assuming this blog was relevant to the original intent of the searcher!). Better queries are possible, but if natural language queries work, isn't it just being anal to insist that more elegant queries be used?

Second, and closely related, there are no failed queries in this list. Everyone came to this blog as a result of a query.  No matter what I might insist makes a good or optimal query, a lot of variation and clutter (unnecessary terms) still produces effective results. One of the techniques I like to use with students is to display a list of queries they propose for a specific type of search. One way to collect their queries is to create a Google form and have them post their query to the form. You can display the results in a spreadsheet without names attached to the queries. That works pretty well because there are almost always queries that will not work. It's informative for students to see how others create queries. Better queries are pretty easy to identify--which I think you can do with the list above. However, poor queries can be effective too, so it's always a good idea to try the queries to see if they connect to the information that was sought.

Third, looking at queries reveals trends. Looking at more than 30 queries shows me what topics brought people here. There's quite a bit of interest in detecting bias, which will prompt me to develop more resources on bias/objectivity in evaluation. If you have a blog, I recommend you unpack the queries that brought readers to you. You may be surprised.

No comments: