As you'll know, I'm working with the smart folk at Adaptive Lab to build a Twitter curation app called Fraggl. We've been a little quiet for a couple of weeks but things have been happening. First up, we have a logo (yay!).
We've also been working through (and incorporating some of the feedback we've had) for some of the options for developing the algorithm. This has confirmed our thinking that using Twitter lists could be a useful way to combine human and algorithmic curation. I really like the idea of enabling the option to plug the algorithm into your whole feed, or into self-curated lists and even into suggested pre-curated lists to see the most popular tweets/content shared by particular groups of people. The challenge here is balancing quantity and quality (too many or too few people/content/links impacts the ability to surface the 'good' stuff). An added dimension here is the frequency that you get updates. The advantage of daily updates of popular links is that the content is fresh and as long as the quality is good enough, higher frequency could work well, but it would need a larger daily input of content to scrape. So the question is whether lists would be enough for that. The advantage of weekly updates is the ability to take in more data, and so in theory to surface better links, but then you may get well shared links that everyone's seen or just old news.
So, in short, we're going to do some experimenting with a week's worth of data from one list, and analyse it based on two or three combinations/weightings of signals including number of favourites, retweets, numbers of followers for that tweeter and so on. We also have some thoughts around some more interesting ways of surfacing great content beyond this but we'll start with what we've got and see where that gets us. Once we have some results well share some thoughts on what next. In the meantime, my thanks to Adaptive Lab, and if you'd like to be kept in the loop with (infrequent) updates and to be a tester when he have a live product, you can sign up here.