Recently, I came across Phrase Nets, a new kind of text visualization, and I thought “wow, this could be really useful to researchers seeking information”. My husband, a researcher seeking information, looked at the same thing and said “wow, this is really useless, it only shows you obvious things”.
A phrase net of Pride and Prejudice with the pattern "X and Y"
The example visualization, reproduced above was a network of the pattern “X and Y” applied to Jane Austen’s Pride and Prejudice, and yes, the names and clusters that show up are obvious. I didn’t gain any new insight from looking at the picture, so why did I think it would be useful in any way?
Automated tools like this, that pick up on so-called obvious information, are easy to dismiss. We say – oh, it can’t tell us anything more than we already know. What we forget is that sometimes we’re looking for exactly this sort of information: stuff that is obvious… but that we don’t know yet.
Returning to the phrase nets example, suppose that the text wasn’t Jane Austen’s Pride and Prejudice, suppose it was the entire collection of journal papers on Natural Language Processing… or Neurobiology… or North American Archeology.. or 17th Century English Politics — some academic discipline. Now suppose that, while reading some book or paper, you came across a researcher’s name mentioned along wtih some claim or advance of interest to you. e.g. “… the idea, advanced by Hearst in 1989 …” and you wondered “who is this? what did they do?”
I imagine you could use a phrase net with a pattern like “Hearst claimed/showed/demonstrated/argued Y” (or however these things are phrased in your field) to learn something useful. Obvious to anyone who actually knows Hearst, of course, but still useful to you.