I didn't really understand what this graph shows. I saw that he was fetching the from and to headers but then what? Does it show how the from contacts and to contacts connect to each other?
Would be great to animate such a graph over time and see how social circles change over the years.
Also I think theres other fascinating data in email. I'd like to run over my inbox archives and for example extract topics of interest that occupied my mind at certain times. It would probably be very clear when I studied what topic, when I was sick, when I was traveling, what women I've been with etc. I even heard somewhere it's possible to predict the chance to get Alzheimers by analyzing the increase or decrease of vocabulary over time. Data mining is a cool subject.
I really hope this isn't the way most people use R. Wouldn't it be better to call out to a python script that pulls in the emails and spits out a CSV, and then load that into R?
This is what I call as R bias. I am regular user of R but I would never use R to prove the point that this can be done in R. How religious is that? Sorry about my R bashing but users of R need to understand that what kind of problems R suppose to solve.
I'm not going to defend this particular set of code or analysis (it's not how I would have done it, and it's a bit silly to use R to wrap a bunch of python code), but the notion that you shouldn't use R for this out of hand is off to me, and just as religious as saying you should always use R.
Use whatever programming language YOU are most productive in. For me, that's often R, even if there's a language that's "better" for a specific task -- it's not necessarily better when it's in my hands. FWIW, there's a much easier way to do this in R as well (libcurl 7.20.0 and later includes smtp, imap, and pop3 protocols, letting you do this without calling out to jython).
There are lots of Turing complete languages. They all can do anything. There's no need for religion about language in any case. Hell, write your killer web app in LOLCODE for all I care.
Doing data massaging is IMHO painful in R and Matlab, and this is often a huge part of the code. RPy looks like it has fairly usable syntax (eg seems to support Python-style arguments rather than nasty compiled strings) so I'm hoping I can get away with doing my scaffolding in Python and calling R stuff when needed as more or less another library....
12 comments
[ 5.5 ms ] story [ 33.2 ms ] threadI didn't really understand what this graph shows. I saw that he was fetching the from and to headers but then what? Does it show how the from contacts and to contacts connect to each other?
Also I think theres other fascinating data in email. I'd like to run over my inbox archives and for example extract topics of interest that occupied my mind at certain times. It would probably be very clear when I studied what topic, when I was sick, when I was traveling, what women I've been with etc. I even heard somewhere it's possible to predict the chance to get Alzheimers by analyzing the increase or decrease of vocabulary over time. Data mining is a cool subject.
I understood he wanted to know what the women he had been with were like - what were their interests, their social circle, etc.
Use whatever programming language YOU are most productive in. For me, that's often R, even if there's a language that's "better" for a specific task -- it's not necessarily better when it's in my hands. FWIW, there's a much easier way to do this in R as well (libcurl 7.20.0 and later includes smtp, imap, and pop3 protocols, letting you do this without calling out to jython).
There are lots of Turing complete languages. They all can do anything. There's no need for religion about language in any case. Hell, write your killer web app in LOLCODE for all I care.