Yale Daily News Online

By: Henry Corrigan-Gibbs (Director of Online Initiatives)

Jump to Comments

In class today I was sitting behind another student who appeared (to the professor, at least) to be furiously typing notes into his laptop.  From my perch behind him, I could see that this kid was browsing the Yale Daily News Web site.

While I was perplexed that he bothered to drag himself out of bed for a 9 a.m. lecture that he wasn’t even going to listen to, I was ecstatic that he chose to spend his time looking at our Web site.

Competition for viewers online is fierce so once we get a reader on our site we try to do everything in our means to keep them from browsing to some other entertaining Web site.  An extensive tagging system is one of the prime ways we’ve come up with to encourage people to explore the site instead of leaving after reading just one popular article.  By providing links to other content that users might be interested in, the site tries to guess what interests each viewer.

For example, when a viewer is reading an article about Richard Levin, they will be given links to related content pages about the Yale Corporation, the campus expansion, and Richard Levin himself.  Once on those topic pages, a viewer can click through to any number of other articles, tags, photos, or videos related to that topic.

Cool, huh?

Conceptually, the tagging system is very simple.  When editors upload articles to the site, they add content tags to each stories by hand.  Once the story has been put online, the content management system does the rest of the work.

One of the problems we ran into with this feature was that the editors refused/forgot/didn’t want to add tags to stories and photos as they upload them to the site.  We quickly realized that if wanted the system to get used, the content management system would have to tag stories automatically.  While still allow editors to tag stories manually, we now tag stories automatically when they’re uploaded by searching through old articles and guessing where a tag might apply.

While we were initially skeptical about having an auto-tag feature, it turned out to work pretty well.  For example, when looking for articles to tag with Admissions, the system finds:

• Elis graduate with no skills — and few prospects
• Ivy admissions prompt frenzy
• Acceptance rates decline across the Ancient Eight

Pretty good.

However, when the auto-tag feature looks for articles to tag with the word Court (as in the New Haven County Superior Court), it doesn’t do as well:

• Appointment of Tony Blair sullies University reputation
• Before Yale, Gentry a ‘finesse player’ on court
• Shin found guilty of faking doctorate

We generally ended up getting more sports articles with the word court (e.g. basketball court, tennis court, etc.) than articles about the legal system.  Refining the tag name a little, by changing Court to Crime, makes the search work much better.

To incorporate the tag system into the user experience, we decided to put lots and lots of links to tags pages all over the place.  Every time the editors post an article, the system automatically links every instance of a tagphrase in the article body to its corresponding tag topic page.

Look familiar?  We got some inspiration from the New York Times’ application of the same concept.