on Textmap.com and Google Sets

Textmap functionality based on entity exploration is difficult if the ‘Entities’ button doesn’t appear on the front page. I’m not sure if this is something that has changed over the past week or two, but the only button at the top of the front page (http://www.textmap.com) is ‘Contact’; only by going to the ‘About’ page (http://www.textmap.com/about.htm) was I able to see two additional buttons appear, ‘Entities’ and ‘Sources’. Also, I don’t understand the basis for ‘positive raw counts vs. negative raw counts’ – so, much as I can appreciate the impressive nature of designing something that trolls through news, aggregates and ultimately interprets data, the lack of a full explanation of statistical significance, (or maybe just my lack in vocabulary) means that I don’t ‘get’ this project. On a higher level, I understand it as a tool for interpreting news and building connections; from the perspective of a user, though, I don’t find the site that easily navigable, (a list of letters beneath each entity type doesn’t exactly encourage browsing, or at least doesn’t encourage browsing that isn’t frustrating!), or intuitively understandable as regards purpose. The language describing each graphic representation isn’t obvious—the site would be much more useful, I think, if there were some sort of indicators of (1) what the data was based on, and (2) why the comparison data is significant. There is no useful glossary or explanation on the site, even on the ‘About’ page.

The Google Sets function is interesting, and I’d be curious to see what data Google draws from (and how) to determine connections, in order to fill out a ‘set’. I tried a few terms – Jens Lekman, Sweden, Acid House Kings, (the first and third Swedish musicians, the second their home country), to see how this would work. It did well, in a sense, as the set predicted was a long list of bands from Sweden and on the same or associated record labels. (A somewhat similar function can be had, although only starting from a single musical entity, at http://audiomap.tuneglue.net/). I don’t see, though, that this sets feature is actually building—or even implying—relationships between the search terms. For instance, when I searched for Richard Nixon + Burma + Lyndon Johnson (also tried Richard Nixon + Myanmar + Lyndon Johnson), the results list of US presidents, with a couple of VPs and ‘Iraq’ near the bottom of the long set. On the up side, it is easy enough to use (by just listing terms in different boxes), and Google even included suggested searches under ‘Examples’ at the bottom of the page.