Brainfood and Mozilla’s Open Innovation Team Kick Off Text Classification Open Source Experiment
Mozilla’s Open Innovation team is beginning a new effort to understand more about motivations and rewards for open source collaboration. Our goal is to expand the number of people for whom open source collaboration is a rewarding activity.
An interesting question is: While the server side benefits from opportunities to work collaboratively, can we explore them further on the client side, beyond browser features and their add-on ecosystems? User interest in “filter bubbles” gives us an opportunity to find out. The new FilterBubbler project provides a platform that helps users experiment with and explore what kind of text they’re seeing on the web. FilterBubbler lets you collaboratively “tag” pages with descriptive labels and then analyze any page you visit to see how similar it is to pages you have already classified.
You could classify content by age or reading-level rating, category like “current events” or “fishing”, or even how much you trust the source like “trustworthy” or “urban legend”. The system doesn’t have any bias and it doesn’t limit the number of tags you apply. Once you build up a set of classifications you can visit any page and the system will show you which classification has the closest statistical match. Just as a web site maintainer develops a general view of the technologies and communities of practice required to make a web site, we will use filter bubble building and sharing to help build client-side understanding.
The project aims to reach users who are motivated to understand and maybe change their information environment. Who want to transform their own “bubble” space and participate in collaborative work, but do not have add-on development skills.
Can the browser help users develop better understanding and control of their media environments? Can we emulate the path to contribution that server-side web development has? Please visit the project and help us find out. FilterBubbler can serve as a jumping off point for all kinds of specific applications that can be built on top of its techniques. Ratings systems, content suggestion, fact checking and many other areas of interest can all use the classifiers and corpora that the FilterBubbler users will be able to generate. We’ll measure our success by looking at user participation in filter bubble data sharing, and by how our work gets adapted and built on by other software projects.
Please find more information on the project, ways to engage and contact points on http://www.filterbubbler.org.