I've noticed some interesting behavior with the docSentiment processing:
"XXX, you're an idiot" gets scored neutral (!)
"XXX, you're stupid" gets scored negative
"XXX you idiot" gets scored neutral (!)
"XXX is an idiot" gets scored negative
Seems rather arbitrary. Will the processing model "learn" like Watson does? Or is this running up against the limitations of our ability to process natural language and all it's nuances?
Please sign in to leave a comment.