posts tagged ‘recommendation systems’

Finding inbox meaning in a time of overload

4 February 2011 | Victoria Bellotti

[contributed post to Inbox Love, produced by 500 Startups] Email is, for many knowledge workers, a habitat: the place where they spend most of their working day online. Indeed it can be thought of as knowledge work’s Grand Central Station as far as information distribution and workflow are concerned. A major part of knowledge worker information overload is trying to manage the influx of email content in terms of prioritizing obligations communicated via email and making sure they can always locate the resources they need within all the content in their inbox. In order to handle the demands being placed upon it, email needs to be far better integrated with its users’ content, communication streams, and productivity tools, and come pre-armed with powerful features to support things like content organization, project planning, workflow, content retrieval, analytics and so on...

It’s time to reap the context-aware harvest

21 September 2010 | Bo Begole

It's ironic that following the invention of the Personal Computer workstation and laptop computers at PARC, researchers would then turn toward making the computer disappear. To most people at the time, having a single “personal” computer was a dream, but Mark Weiser and many others envisioned that we’d soon all have more than just one personal computer in our lives... Today, context awareness isn't about devices and location - it's about people getting things done.

Want to be retweeted? Add URLs to your tweets!

16 August 2010 | Lichan Hong

In a previous post, I described a recent study in which we found that including hashtags in a tweet enhances the retweetability of the tweet. In this post, I'll focus on another factor that might affect retweetability: the URL.

Experts’ domain knowledge improves automated recommendations

9 November 2009 | Ellen Isaacs

It’s almost eerie how well the music website Pandora recommends music based on just one example of a favorite song. It does so by relying on human experts to characterize songs based on a large and musically sophisticated set of characteristics (melody, harmony, rhythm, orchestration, etc.). This approach -- of using human expertise to develop a rich set of attributes that deeply capture the essence of an item -- could be adapted to greatly improve the recommendations currently being offered in other domains, such as news, movies, hotels and so on. In our enthusiasm to develop automated recommendation systems, Pandora reminds us of the value of incorporating the intelligence of domain experts into the process.

Can a recommender choose what’s best?

7 July 2009 | Ellen Isaacs

While doing a user experience evaluation of a mobile recommendation system, I noticed how the list-based design implied that there was a single best choice, with 'goodness' decreasing in even increments. This made me wonder whether it makes sense to try to identify the one Best option when you're choosing among options that vary across many dimensions. Making choices might be easier if we assume that there is rarely one single Best.