2
What problem is Gloe solving?
As more devices have access to the Web and more people contribute content to the Webwe are suffering from
information overflow
. This problem is exacerbated on smalldevices with limited screens, input controls, and bandwidth. To solve this problem, and topiggy-back on the success of the Web browser interface including the upcoming HTML 5geolocation support, we designed a geo-aware market for local Web content that givesincentives to (and makes it very low-effort for) both consumers and providers tocontribute and evaluate content. Our solution allows non-intrusive plug-ins to be writtenfor existing browsers to enhance the local search experience without taking away fromthe well-known, well-proven browser interface. Finally, our service is designed to work well on limited bandwidth devices with small visual real-estate by automatically bubblingup the most relevant and popular content in a filtered, ranked list of recommendations.
How does Gloe work?
Gloe is designed around a massively parallel backend where data are split across smallregions that can be efficiently queried for all information in a geographic area. Thismeans that local search is very fast at the expense of global search not being supported inreal-time. Each recommendation by a user is recorded as the URL, title, latitude-longitude coordinate, and an arbitrary tag. The recommendation feeds are constructed bysimply aggregating over recommendations for the same URL within a region determinedby a latitude-longitude epicenter and a radius. To ensure high-quality contributions and toinfer popularity more accurately (i.e. to give the right incentives to both consumers andproviders of information to contribute), all users in the system are granted arecommendation budget. This budget is used when making recommendations in thesystem. The users specify how much of the budget to use on each recommendation,which can be compared to rating an item on, e.g. a 5 star scale in a traditionalrecommendation system. When running out of budget, no more recommendations can bemade. To simplify the interface to users, a recommendation by default spends 1 creditand to, e.g., spend two credits the user could recommend the Web page twice. Theaggregate budget spending on a URL in a geographic region is the ultimate metric forpopularity in Gloe, and it is what is used to rank the recommendations. Informationproviders don’t necessarily need an incentive to contribute and may even be willing topay to contribute because they typically want to promote their own content. However,information consumers may need to be given an explicit incentive to evaluate andrecommend content, for example, having budgets be reset periodically in a use-it-or-lose-it mechanism.
How can I use Gloe?
Gloe can be used today for all the use cases described next. The service is designed as anopen-ended general purpose Cloud service that can be used on many different platforms.Currently we have clients for all major desktop Web browsers (Firefox, Safari, Opera, IE,Google Chrome). We have native apps for Android and RIM/BlackBerry, and the servicehas also been tested to work well from iPhones (supports HTML 5 to obtain the GPSlocation) and iPAQs using their respective Web browsers.
Leave a Comment