Subscribe to Site Feed

what's new f.a.q about us feedback

history women portraits mythology

arts music culture paintings

tags sitemap timeline index a-z

pictures! stamps erotica biographies

gandhi tribals temples religions

 

Best of AnthoBLOGy • Unripe Revolution • Rooster's Dharma • Don't Know Jack • No Love for Condi • Blogger's Block • Father of the Bride • TinTin's Diary - I • TinTin's Diary II • Hate Bollywood • Child Labor • M.F.Husain Guilty •Marathi & Konkani • Artist's Daughter • India's First IT Guru

Computing, Libraries, Tennis, India & other interests of Vikas Kamat

Computing with Tags

I am happy to add a new "Browse by Tags" feature to Kamat's Potpourri. We have internally supported tags from 1997 (we called them CrowWords, a mock of keywords, after my company CyberCrow), but they were designed for machine-learning rather than for humans. Anyway, the addition of tags in addition to keywords and crowwords gives us a rich content platform classify and organize the website. It is indeed fascinating to see the disparate content items come together. Hope you like 'em. Examples: Hampi, Hindusim A Note About Tag Generation It is wonderful if you have the luxury of human-tagged contents (like Flickr, Technorati or Delicious have). But what if your content is not already tagged? That's when Keyword Extraction comes into play. We had implemented a rudimentary extractor for our -- if a letter was capitalized, it was given a weight and a tag was started. I don't have any research or study on how effective it has been. Then Yahoo introduced an Extraction API. In my experience, it is a wonderful service (if you can tolerate the 503 s it seems to throw often), and we have supplemented generated tags whenever human-provided tags were not available. I had to make much tuning, because most of the tags Yahoo suggested had "kamat" or "india" in them. But this automation allowed us to gather some 192,000 unique tags! After gathering the tags, I build a matrix to view the tag density. This exposes both the strength and weakness of our contents, and helps in generation of Tag Clouds. Generation of Tag Clouds The coma delimited tags are transposed into rows using SQL and are ranked based on repetitions. The physical proximity is weighted in. Having all this infrastructure, the generation of a Tag Cloud is then simply querying the database with various SELECT clasues. A Sample Tag Cloud: beauty - blog - diwali - games - goddess - huntress - india - lost - photo - photographs - potpourri - wheels - zoo - accidental art - ant - aperture - beholder - child delivery - color yellow - colors - cricket - cultural anthropology - destiny - friends - god - illustrations - jewelry - kamasutra - mirror project - mysore - opaque - photo album - prehistoric - professions - refugee - seductress - spokes - torn - tri (Comments Disabled for Now. Sorry!) First Written: Wednesday, October 4, 2006 Last Modified: 10/6/2006 2:55:25 AM Tags: tagclouds

Browse More Entries

About Me:

• My Home Page
• Contact me
• Kamat Family Album

Powered  NOT by Blogger or MovabaleType or WordPress, but by SimplyBlog, a software I wrote to create blogs. See details of implementation or download SimplyBlog.