« first day (546 days earlier)      last day (2186 days later) » 

2:56 AM
I'm lonely right now :(
 
3:08 AM
Hey @rolfl
Welcome to my little corner of the web!
Kinda
 
There, s'not so quiet anymore.
 
@rolfl That's a good thing
 
need a 3rd monitor.
 
Haha
I don't have one, so I'll have to keep switching tabs between this chatroom and the Code Review chatroom
@rolfl Now that you have an account here, you can upvote me!
 
@syb0rg - I have used RPi a few times as it answered some of my questions... my commitment is to CR though.... and the rest of my IRL.
Don't expect me to become a regular here. ... ;-)
 
3:22 AM
@rolfl I'm just glad you joined. As long as you come and visit every so often I will be more than happy.
 
4:21 AM
@Corbin Well, we need to get more people integrated into this chat room in my opinion.
 
All the starred messages are so depressingly old >.<
 
@Corbin I just starred the message from @rolfl
It's a movement!
 
:p
 
@Corbin You don't have any asked questions or answers on this site?
 
nope. but i've been a member for a year and 3 months... lol
my damn SE account addiction
i actually didn't know that there is a rpi SE until today
and apparently some point approximately 1 yr 3 months ago
 
4:24 AM
Haha, awesome.
@Corbin Would you be willing to upvote some of my answers on this site if I look through some of yours on Code Review and upvote them?
 
hrmm, not sure how that bodes in terms of SE ethics... but sure, assuming you mean only answers that legitimately warrant upvotes. i can look through a bit later.
at the moment, i'm stuck working on a project :(
 
@Corbin For?
 
also, i think there's an upvote streak reverser that undoes mass upvotes
twitter sentiment analysis
 
@Corbin I'm only upvoting around 5 right now for you. After I read through them of coarse.
 
:)
 
4:30 AM
@Corbin So explain this analysis to me.
 
basically it's a term project for my advanced information retrieval class
 
And what specifically are you retrieving?
 
we had to pick a topic, do a 30 minute presentation on it, and create a program that explores the topic. basically we have to find a few research papers, present a summary of them, and recreate a watered down version
well, typically IR focuses on retrieving documents based on search terms (think good)
but this is a bit of an AI/IR combo that focuses on taking documents (or in this case tweets)
determining who the speaker is, what the target is, and how the speaker feels about the target
for example, if someone tweets "@justinbieber is a loser!" then they obviously do not have very fond feelings towards justin bieber
there's a lot of ways to go about sentiment analysis. the approach i'm taking is an n-gram/neural net hybrid
an n-gram is just a collection of n words. so like "The dog ran" 1-grams would be the, dog and ran. 2-grams would be "the dog" and "dog ran". so on
a neural net is a computer simulation of an organic brain (sort of)
it has connected neurons that form a bit of a mesh. information goes in, data comes out. the complicated part is training the neural net to do what you want and do it accurately
and yeah....
 
@Corbin Interesting. So you are doing web mining on Twitter for the emotions that tweets display to people. And to do that you are interpreting each word individually to determine the emotion that is being conveyed.
 
i have a very unpleasant night ahead of me since i've gotten about 3 minutes worth of presentation, 10% of the programming part done, and it's due in 12 hours x.x
sort of
 
4:34 AM
@Corbin Procrastination?
 
you can't actually consider words individually and get a good accuracy
for example, image "why all the hate against ___? I think it's fine!"
hate implies a strong negative feeling, and fine implies a neutral feeling
so on the surface that would appear negative
but really it's more towards positive since it's defending something
that's why it's not as simple as just summing up a bunch of word weights
 
@Corbin So how do you go about associating those words with what you are describing?
 
there has to be a natural language processing/machine learning component to it
basically you have to make a program that's smart enough to recognize that in that context, hate does not actually mean hate
overall, it's quite difficult. the academic community seems to be getting around 60-70% accuracy at the moment
what complicates it even more is that people only agree on tweet sentiment 79% of the time
 
@Corbin And your accuracy?
(rough estimate if you have to)
 
well... 0% for the moment since i don't have that part done yet
but my guess is... meh... probably like 40%
maybe worse
in 12 hours i can't get too sophisticated
it's going to be very rough
lol
 
4:38 AM
@Corbin Does the class require a threshold accuracy, or is it more of a proof-of-concept?
 
it's a proof of concept
it's just me doing this project
other people are doing different projects
for example, one student is working on something similar to google's distributed file systema
and someone is working on identifying themes and motifs within novels
 
@Corbin Sounds interesting.
Well, I'll let you get back to it. I have to go.
 
it sounds more interesting than it is :(
see ya
 
 
18 hours later…
10:29 PM
@@psp Hey psp. Would you mind coming to [chat]? Or at the very least, read the site faq as I believe you are not understanding the purpose of Stack Exchange. Thanks!
 
10:50 PM
@Jivings Hi
Aww, are you not on your computer?
 

« first day (546 days earlier)      last day (2186 days later) »