Comment by tokenadult | Hacker News Reader

Comment by tokenadult | Hacker News Reader

This interesting set of maps is an example of how presentation can triumph over content. The pretty maps total to not a huge amount of information on dialect differences in the United States, a topic that has been posted about here on HN before, but their sheer prettiness prompted me to share the link to my Facebook wall, whereupon Facebook showed me on my home page that five of my other friends have already shared the same link. It's often easier for pretty pictures to go viral than more informative paragraphs of text.

The underlying survey that gathered the data

http://www4.uwm.edu/FLL/linguistics/dialect/

is, I think, something we have also discussed before here on HN. The link from the survey home page that is supposed to lead to the faculty webpage of the principal investigator appears to be a dead link. He is now at a new university,

http://www.mml.cam.ac.uk/dtal/staff/bv230/

still working on linguistics research.

I disagree about the quality of the information presented herein. As someone whose day job involves recording sound for film (necessarily including lots of dialog) and who is not American, regionalisms in accents are both challenging and fascinating for me, and a frequent topic of discussion between myself and the actors I work with. On smaller films there isn't usually the budget for a dialog coach and since I listen to people for a living that aspect of production often falls in my lap.

I found these maps extremely informative and and they helped to clarify some perplexing edge cases (eg people from Missouri sound southern in general, but on some words they exhibit marked differences from their near neighbors).

Not to put words in his mouth but, I think the GP was referring to the information density rather than the information quality itself. Likely as the comparison he was making was to the presentation quality.

I think you are both capable of being correct... but then I must think that because I agree with both of you :)

While I'm not an expert in the topic, the visualizations would seem to be drawn from a fairly substantial data set--over 30K respondents. http://www4.uwm.edu/FLL/linguistics/dialect/staticmaps/q_all...

It's really unfortunate that the data is presented at such a high level of aggregation.

It would be so much better to have individual level data (perhaps with personal info such as age, etc. removed) so other people could do their own analysis (e.g. clustering).

I don't want this to be interpreted as criticism of Bert Vaux and his colleagues: there is incentive in academia to not share data, since you get credited for the uniqueness of your work, and get no credit for collecting data used by others. I just hope the situation changes.

Reply on news.ycombinator.com