Real-time search, noisiness and influence
There’s a really fascinating post over at TechCrunch today by Mary Hodder, someone who’s been working in ‘live search’ – what we now call the real-time web – for some time.
The article’s definitely worth reading in its entirety, but I wanted to highlight some of the difficulties with real-time conversations that she mentions. A great example is the Michael Jackson Tweet-splosion; if you’re taking a purely search-based view, what do you search for? “MJ”? “Michael”? “King of Pop”? As Mary says, that’s a relatively easy example!
More interestingly is the comments Mary makes about authority. How do you measure authority online? Well, as part of my initial PhD research I looked at various web-structure algorithms (yes, including PageRank) and how you might exploit them along with semantic information to gain a true understanding of the importance of an article.
This research is rooted in scientific publications, in fact; we can learn a lot from the relatively ‘clean’ case of scientific paper citations, although the language used on the web is about a thousand times more interesting. (And, thus, a thousand times harder to process.)
If I told you how we actually track influence, of course, I’d have to kill you. But check out Mary’s article, it’s great food for thought.
Recent Comments