Follow Surfarama on Twitter | Subscribe to RSS Feed | Subscribe via Email

Tagalyzer

OK, so here’s a bit of an experiment…

Try clicking this link.

What you’ll see is a breakdown of recent news, blog posts, bookmarks, and pictures related to the Wikipedia article on hackers, achieved by running a content analysis of the wikipedia page and searching for related content. Cool huh?

Tagalyzer.com takes a referring url and applies the Yahoo! content analysis API to identify the important keywords, and then uses those keywords to poll the Yahoo! News API for recent news, the Technorati API for blog posts, del.icio.us for bookmarks, and of course Flickr for pictures.

If you just link to the main page it will pick up your referring url automatically, but you can also grab the TagIt! bookmarklet (drag it to your browser links bar) for returning related content for any page you’re browsing. Or, thirdly, you can craft a url like “tagalyzer.com/?referer=SOME_URL” and it will take “SOME_URL” as the referer. This is what I did above and it would be useful for building into your blog template for example if you wanted to add a link for each of your post permalinks.

Actually that quick how to also kinda summarizes the three use cases I see:

  • Any page just links to tagalyzer.com and click throughs return related content from the referring page. Simple
  • You craft a url with the referrer variable and add it to your blog template for each post, as I have here.
  • You use the bookmarklet to discover related content as you traverse the interweb. Nice.

Of course there is a catch…it does require the referring page to be quite topic specific to get some good results. A link referring the home page of someones personal blog for instance would probably comes up with some pretty random content due to the generally non specific nature of most personal blogs…for this reason you’d be better off linking from your permalinked pages so the analysis focusses on the topic of a specific blog post.

Anyway, keen for any feedback. Useful? Buggy?

3 Responses to “Tagalyzer”

  1. BrianHamlin Says:
    July 20th, 2006 at 5:22 pm

    Hey Charles- I just returned from MashupCamp II. Got to meet a bunch of people behind the websites, and took a chance to look through ProgrammableWeb a bit more. I found the Tagalyzer.

    Its use of the APIs is clear and beyond any of the others in this category that I saw. However, I guess it is the Yahoo API to do the initial analysis that is wanting. The results were not good in general (with my limited try outs), and using only one tag at a time to fetch related links gives _way_ too broad a result.

    Something needs some sharpening up. Since I am relatively new to server side programming, I am naturally curious as to the tools you used to create the thing, as well.

    all for now, and best regards from Vancouver, BC
    -Brian

  2. BrianHamlin Says:
    July 20th, 2006 at 5:24 pm

    btw- I’ll be at WordCamp this coming Aug5 in SF. I use Wordpress as well…

  3. Charles Says:
    July 23rd, 2006 at 7:02 pm

    Hey Brian,

    Thanks for your comment. You’ve hit on exactly the reason that the tagalyzer experiment hasn’t been taken any further…the Y! content anlysis provides insight into what Y! *thinks* is important (as a search engine), but not necessarily what we would take a way from a particular page. I kinda hoped that keyword analysis would be a good proxy for automated ‘tagging’, but it doesn’t really seem so.

    The site itself is really pretty straight forward…taking a referring URL and running it past the Y! content anlysis API to identify the important keywords on each page. These are displayed in order of importance (according to Y!), and each triggers the corresponding search in each category. All pretty basic, and wrapped in a little ajax to smooth the UI a bit.

    If you;ve got any ideas about how this could all be improved I’d love to hear them.

    Cheers

    Charles