I'll need to talk about corpus linguistics in the upcoming Ten Hundred keyboard video, which reminded me of a 1981 book I came across in grad school. It attempts to find personality markers based on some fairly simplistic statistical analysis of people's speech. Not particularly successfully, I think, but I was working on a system to synthesize language with emotional cues and this was one of the only sources of quantitative data I could find. It even spends a chapter analyzing the Watergate transcripts using the same tools! My final project for that class had sliders for how schizophrenic, delusional, depressive, compulsive and Richard Nixon the generated text was supposed to be.
Anyway, a copy just arrived so I can include it in the video.
=> More informations about this toot | View the thread | More toots from attoparsec@clacks.link
text/geminiThis content has been proxied by September (3851b).