I had a weekend full of mind-flexing! It's enough for a whole bunch of meditative blog-postery so that is what it will be. Just not yet.

In short, though: the lady's university had a departmental conference which she had helped to organize. I presented a general overview of my work at said conference; this is especially notable as I am a student of the Life Sciences and this was a conference of German language and literature. It was, however, intended to be interdisciplinary by design. It succeeded quite well and I have some fresh ideas to think about. There was an especially nice talk about repetition in music. See the following example, Steve Reich's It's Gonna Rain:

I've also been trying out a few artsy games: Antichamber and Starseed Pilgrim. They're elderly in internet years but young n' fresh to me. Thoughts later.

I don't normally feel unnerved about the omnipresent social-networking morass persistently enveloping the modern internet. It's clearly a social phenomenon and an admittedly interesting one at that. Even so, I can't help but feel a tad creeped out when. say, I edit a company's new Wikipedia page and said company's founder looks up my Linkedin profile a few hours later. He doesn't even leave any snide comments or anything, he just looks at it, a feature seemingly unique to that particular network. It's not easy to find my LI profile from my WP one without some Internet Detective work so it really just comes down to a big "why?".

Why do I care who feels like being strange on, god forbid, the internet?
Why do they go and do it anyway?

(Of course, the Wikipedia Internet Detective also felt like looking up one of my email addresses and sending me a passive-aggressive nastygram. Maybe a better question is "why do people get offended so quickly on the internet?")

Here's my glitch for the day.

We talked about the Peak-end rule in the development class this evening. It's an interesting heuristic, if perhaps an oversimplification. We discussed it in the context of giving presentations but I'm curious about how it applies to other presentation-like experiences or even whether a collective group of people tends to define the "peak" of a shared experience in an identical manner.

Thar's gels in them thar hills

Text mining is a rather unkind metaphor. In the life sciences, it refers to how we can sort through all the published data out there and extract broad conclusions from the aggregate. This does make the rhetorical assumption that most of that text is some kind of stone to be blasted away until the truth emerges, gleaming in the sun.

In practice, things are never that simple, of course. I think I've mentioned here before how some of the major science publishers have only just begun opening their archives to text mining. There's also the issue of images: current software can usually handle text without serious issues but extracting meaningful data from figures is conceptually problematic. Scientists just aren't consistent when it comes to presenting their data. That's a good thing, really, as it they often have to focus on different aspects of their results; nature (or Nature, for that matter) just isn't always as consistent as we'd like it to be.

A paper posted to arXiv recently by Kuhn et al proposes one strategy for extracting data from images in scientific papers. They focused on gel images. These kinds of figures are great because they're generally just photos with sets of horizontal bands in them.  The placement of the bands determines their relative size, so as long as a size standard is present, there's one bit of easy data already. The tricky part is knowing what's actually present on the gel. Even when we know that, the gel images are often too data-rich to tease apart every apparent result. As the authors say, "...the text rarely mentions all these relations in an explicit way, and the image is therefore the only accessible source."

These folks used a straightforward approach to break down gels into usable data - check out the paper for details. They're fond of the optical character recognition in MS Office 2003 for handling text. The gel segment recognition needed a machine learning-based approach and random-forest classifiers. Assigning relations to those gel images is much trickier so the authors had to use the ol' Human Touch in their code.

So how well does it work? Not terribly well yet, at least because it's still incomplete. The issue of inconsistent labeling remains; the authors' approach works passably as long as figures are neatly labeled with gene or protein names. This kind of approach and others like it may eventually mean authors have to consider image mining when designing figures. They could, of course, just have fun with the image mining and write out all their labels by hand.