Thursday, October 28, 2010

Automatic Searching for New Papers



I'm always on the lookout for new scientific articles that I can use to inform my work, and cite in my books and papers. But what does being on the lookout entail?

I get papers in the following ways:

  1. People recommend them to me
  2. I read about them in secondary sources (popular science books, magazines, and blogs)
  3. I review journals periodically and try to find relevant papers. 

I'm sure I'm not getting all of the papers I'd like. These methods are haphazard and prone to missing large subfields, because I simply don't know to look in them.

When I find a scientific article I like, I download the PDF and put it in my "articles" directory. They're all in there. Hundreds of them. Overall, they give a pretty good picture of what I'm interested in.

What I'd like is a bot that can analyze these papers, and use them to find papers that are new (to me).
I also enter the books I read into google books. See my blog post on this topic at
http://jimdavies.blogspot.com/2009/08/how-to-search-your-paper-books-as.html

It's all available. I would love to get recommendations from an AI.

To go one step further, it could use collaborative filtering (like Amazon.com does) to recommend papers to me that others similar to me have downloaded, cited, or otherwise shown an interest in.

Pictured: Tungsten Rods



Bookmark and Share

1 comment:

Daniel said...

This doesn't relate to your point, but I have found it very helpful to get a few journals' tables of contents emailed to me when the issues come out. That way I can process it just like any other email in my inbox. Read some good new TICS papers today because of it. And you know about keyword alerts in science direct, right?

Also, did you see that gmail just added a feature to make "high priority" messages come to the top? Presumably based on the contents. No idea how it works, but it could be the start of personal AI filters.