I live with Jack and the results are extremely interesting.
People could learn so much about themselves and their relationships just by running some basic sentiment analysis on the conversations they have in Messenger.
I ran this and found some really interesting data to back the context around relationships I have with friends. Really interesting to see male vs female, college vs work vs social, new best friends vs old best friends, etc.
This is super neat! For usability reasons, I think it might be good to add some labels. For example, on the two "by Year" graphs, there are two lines, and I have no idea what they mean.
On sentiment analysis. I have a friend who posted a lot of status messages that indicated mood. Often on top of the world, often raging about someone or something. I took the time to put 1 year worth of comments into a spreadsheet and gave them a 1-5 score for what I thought her mood was. I wanted to run an Fourier analysis on this to see if a roughly 28 day signal was apparent, but the data was sparse and I didn't feel like finding a way to do it. Can you do that? Can you do it for other people and not just yourself? That would cause quite a stir. FWIW the person I wanted to analyze is probably bipolar, so her posts were always very extreme and that's what got me thinking. My interest was purely science (and humor) but others might want to use/abuse such information.
One thing I have noticed with FB Messages is that if you have a lot of messages (about 410k messages over the span of 3-4 years) its really hard / impossible to actually get all the messages back to the beginning. When I try to download my FB data it actually just leaves the conversation blank with 0 messages. (In the past it would dump everything it could get leaving it incomplete). I have also tried getting the messages via the Graph API but have hit the same limits that after x time it would just timeout and not get any messages. Which really sucks because I wanted to do some analysis and archiving of the messages.
22 comments
[ 4.3 ms ] story [ 47.4 ms ] threadPeople could learn so much about themselves and their relationships just by running some basic sentiment analysis on the conversations they have in Messenger.
OK, I get that I could probably run it in a Debian VM too I am sure. Just a pet peeve of the assumptions around operating system monoculture.
1. it doesn't really work with non-English messages or am I mistaken?
2. I changed my name recently on Facebook and now I'm my own best friend, nice :)
And right, that is kind of a known bug. I discovered earlier today when a friend ran it.
Thank you for the feedback!
http://www.wolframalpha.com/input/?i=facebook%20report#_=_
[disclaimer: I am one of the authors of the graphing library used. So neat to see it in action!]