Hi,
We've been investigating divisi2 for mood/sentiment analysis. I
understand that we're on conceptnet4 with the current version of
divisi2?
At this point we've only been doing the simple check of words and
their relation to good and bad concepts.
Next we're going to try the happy<->sad example shown here:
http://csc.media.mit.edu/docs/divisi2/tutorial_category.html
But expand it to multiple vectors, but what is a good strategy for
this? Do we need to come up with words on opposite ends of multiple
sentiment vectors?
(like the happy<->sad example where you set happy=+1 and sad=-1 and
see where your word falls on the scale)
Or should we instead come up with some sentiment concepts and build
related categories from these?
For example:
[anger, loathing, rage]
[fear, uneasiness, trepidation]
[sad, sadness, grief, mourning]
[happy, elated]
Has anyone tried using conceptnet to get mood/sentiment? It seems to
be working but I'm wondering if anyone has any ideas.
thanks.
--
-robert