Reflect: Linking protein names to rich content, as you read

Continuing with my current fascination with cool ways to enhance the experience of searching and engaging the literature online…

Deepak Singh at business|bytes|genes|molecules tells us about Reflect, a new tool for adding value to online articles:

Essentially, Reflect is an entity extraction engine with a specific purpose, recognizing molecular entities, both compounds and proteins. I have spoken at length about the value of entity extraction, and the availability of a service like Reflect just shows you how useful something like this can be. Using either the Reflect website where you can enter a URL, or the Firefox plugin, you can use extract molecular entities on a webpage quite easily. The service highlights recognized entities, and using your mouse you can get additional details as shown in the screenshot below.

Reflect identifies the names of proteins/genes or small molecules that appear in a body of text and generates a live link to a floating window containing informations (and further linkage) about that entity.

You can check it out at the Reflect website.

I had a lot of fun plugging in blog posts from Ouroboros and seeing what Reflect thought of them. It did a great job with genes and a fairly good job with small molecules, though the higher false positive rate in the latter case was a little disappointing (identifying words as small molecules that weren’t, and linking to things that aren’t small molecules at all, like the word “reset” that appears in a graph about something else).

(P.S.: By the way, Reflect was the winner of the Elsevier Grand Challenge, and its development may have been motivated by the incentive of the prize. In light of that, I just want to clarify that I still think Elsevier is the devil.)

Update: The developers of Reflect have a preprint up at Nature Precedings. Thanks to Hilary Spencer for the heads-up.