tag:blogger.com,1999:blog-8929346053949579231.post5512490602427040349..comments2024-03-23T00:59:24.057-04:00Comments on Sapping Attention: First PrincipalsBenhttp://www.blogger.com/profile/04856020368342677253noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-8929346053949579231.post-64864595179975320302010-12-09T11:08:57.265-05:002010-12-09T11:08:57.265-05:001. This time, I'm just using the 100 most comm...1. This time, I'm just using the 100 most common words for PCA. I've been thinking about using that spread, though, or maybe finding a different index of the dissimilarity of words based on their distribution across books. For choosing a biological/political distinction, I'm not totally sure--either I'll have catalog data by then and could just move out from the words most different between the two call numbers; or I'll just intuit some biological words and then use searches to find words that are particularly closely tied, something like I did for scientific method except actually pulling the words out of the last cloud.<br /><br />2. Nothing. PCA is not choosing categories that actually map on to semantic ones. But sometimes you can intuit that they're close because so many of the patterns that structure authorial word choice are semantic ones. So provisionally, I'm just using some to pretend the results make semantic sense. But in fact, they'll never actually correspond to meaning all _that_ well.Benhttps://www.blogger.com/profile/04856020368342677253noreply@blogger.comtag:blogger.com,1999:blog-8929346053949579231.post-73645570398572894342010-12-09T08:51:34.563-05:002010-12-09T08:51:34.563-05:00Two questions:
1. How did you decide which words...Two questions:<br /><br />1. How did you decide which words to use for PCA? Did they come from that bookcount v. wordcount spread? And when you move to a biological/political distinction, how would you choose then?<br /><br />2. What makes "the" narrative rather than "normative"? <br /><br />This is very cool stuff. I'm looking forward to seeing you refine and elaborate it.Jamie https://www.blogger.com/profile/13542022273476075921noreply@blogger.comtag:blogger.com,1999:blog-8929346053949579231.post-86604284394252333092010-12-08T23:31:38.649-05:002010-12-08T23:31:38.649-05:00this is all pretty interesting stuff, but i say it...this is all pretty interesting stuff, but i say it's a red herring and what you all should really be pursuing is better meta-data if you can get it. <br /><br />what's the plan on that? are their cooperative database projects you could hook up with?Anonymousnoreply@blogger.com