The sentiment classifier is not suitable for German documents. Nearly 90% of all the documents in the news-de collection in watson-discovery are classified "neutral". If I use other classifier-services I get tremendously different percentages. (e.g. use the 'summarizebot.com' sentiment-classifier and you will get about 90% classified as either positive or negative.
Why is it useful?
|Who would benefit from this IDEA?||companies dealing with German documents|
How should it work?