I tried something similar for working out discourse
prosodies. Many words are subjective and can't be
classified, and as Martin said, context is very important.
Some made up examples:
"accident" would probably be negative
"I came across this by accident" is neutral
"decline":
"in profits" - negative
"in unemployment rate" - positive
It could probably be modelled using some sort of fuzzy
values, but they're too subjective and volatile, and
context-specific for there to be a database.
Oliver
This archive was generated by hypermail 2b29 : Fri Apr 15 2005 - 12:18:57 MET DST