Recency and Interestingness in Search

I was looking at the popular searches, as ranked on Google trends. Almost instantly one thing caught my attention -- a whole bunch of these search terms are those related to recent events or news. This should not be too surprising at all and it has also been well studied in the literature [1].

But what really struck me was that when we look at the search results, as ranked by any of the search engines -- the important factor there is relevance and PageRank (or the variant thereof). Google Trends itself shows News results, Blog Search results and Web results. News search results and blog results show what the most recent item are: its all about what is happening right now.

To me, there seems to be a gap that we are not focusing enough on. A common reason why a user searches for a terms like "Ingrid Betancourt" is to find out and read more about some recent events that may be in the News. For example, a user who is unfamiliar with the history of Betacourt's struggles might want to look up her background on Wikipedia, but then might be more keen on investigating the recent stories and the reason why her name is in the News. Once that is done, the interest shifts in knowing what people have to say about this event.

However, none of the search results are really helpful to allow exploring what users are really looking for -- which is interesting stories and blog posts about recent events. What I mean by this is we need a 100 different techmemes constructed on the fly for popular search queries, at the very least. Alongside there can be results from Wikipedia and other sources that might inform the user about the background of the search. However, for most part the links and snippets shown here should really be the ones that can be ranked as interesting stories in a techmeme like fashion: not purely on the basis of relevance alone, not only on the basis of time (or a list of articles since the event that triggered the rise in the search volumes) but a compilation of recent news stories that are interesting.

I'm adding this to my "One more interesting thing to hack up" list.

[1] Why We Search: Visualizing and Predicting User Behavior, Eytan Adar, Daniel Weld, Brian Bershad, and Steven Gribble [Link]

Comments

[Shafqat]

Great post - you're absolutely right. I feel like you are tapped into NewsCred offices and listening to our conversations ;-) Or maybe we are listening to yours, because this is something we discuss a lot nowadays! We even specifically mentioned the FARC hostage isssue! I think news 'events' require content/destination sites created on the fly, but with meaningful content. Its something we are actively working on - we'll get in touch with you to get some feedback. Great blog - good luck!