Sunday, August 08, 2010

Tracking Research

For some time now, I have been thinking about an issue. One has a technique, a niche problem or whatever in mind. An example is the technique of Cuckoo Hashing. Someone writes a paper on it, some implement it, some extend it, yet others find ways to apply it in different areas of CS, write wikipedia articles for wider consumption, or discuss it in books, tutorials etc. How do we track the full impact of this? Citations and citation counts don't quite work (nothing more frustrating for an author than finding their paper cited within a long list of papers on a topic with no specific reason). One can start a ``curated'' web site where one lists most relevant of related work and write a commentary. This has been done nicely for say smoothed analysis or compressed sensing. But ultimately, it will be great if the community can collaborate and develop such sites organically.

In that spirit, Graham Cormode and I have set up a "wiki site" on the Count-Min sketch. This is a simple technique that ultimately proves useful for efficiently estimating many of the statistics one cares about on data streams, and we have been surprised by its applications in machine learning to NLP, computer games, security and beyond, and also by its various implementations in systems to hardware. The site has some structure (including FAQ) and we have populated it some. Sites like this are not much fun unless other researchers contribute -- papers, plots, comments --- whatever to extend, enhance or detract. If you want to modify this site, please email (, and we will add you as a collaborator.



Anonymous Bob L. Sturm said...

Hi Slice,

I have liked the idea of a "curated" website slash live research blog that I have created one myself:

I am attempting to mimic the "Boing Boing model", where there are many contributors to the blog in my field (audio and music signal processing with a particular focus on sparse and atomic representations), fused with a kind of web-based paper reading group. There are so many papers that I want to read yet don't have the time to read and contextualize. So far, I am the only one posting --- and trying to do at least one a day to motivate me to get through my pile. And then, when I return to a paper, I can search my blog to see what I have said, and the thoughts that I had.

10:59 AM  
Anonymous Anonymous said...

Hi Bob,

I checked the blog, you seem to be doing a terrific job of reading and writing about papers. Hope others join in.

What I really wanted for a project was mendeley (PDF commenting)+cloud (collaboration on the web, with a sharing model)....

-- Metoo

2:09 AM  

Post a Comment

<< Home