reCAPTCHA

The other day a friend told me about reCAPTCHA, a CAPTCHA-based program (attempting to block spam bots) with an interesting twist. The developers (CMU CS students) are combining a web-scale CAPTCHA solution with an optical character recognition (OCR) system. Since OCR systems are not perfect, they have trouble recognizing some scanned words that humans can typically recognize. By combining known and unknown words in their reCAPTCHA tests, they can simultaneously sort humans from bots and enable an Internet-scale crowdsourced OCR solution. They're using the human-provided text answers to help the Internet Archive project with their digitization efforts.

Recaptcha and is available for use on your own site, and they have plugins to make it even easier to add to Wordpress or Mediawiki based sites.

Expert Forums and Crowdsourcing

This weekend I had the pleasure of meeting Rob and Heather, CEO and Manager respectively of mp3car.com. They and a small support staff operate this popular community forum and store focused on vehicle audio and computing systems. We chatted about all manner of social web site improvements, including community feedback, member rankings, internationalization, text mining, micro-economies and semantically-backed sites. Many forum sites like theirs (e.g. his friend's site BoardGameGeek) are built on the shoulders of a community of subject-matter experts. Once a critical mass is achieved, the community can take on a life of its own. For such sites to succeed, operators need to be attentive to the community and continuously improve the quality and interactivity of the site. Sites like these are fertile grounds for conceiving of and building specific crowdsourcing features to address one or more of the above issues. Rob seemed keen on the topic of our class, and even offered to drop by one evening for a guest lecture.