Sentiment Analysis for Dutch

We are interested in creating systems that automatically discern between facts and opinions in written text. Put simply, we do this by assigning scores to words. For example, bad = –1.0 and good = +1.0. We can then estimate the average score or "polarity" of a sentence. For details, see our paper.

We are currently processing thousands of Dutch words and you can help us out by tagging some of them. It's actually quite fun! Moreover, we give back results in the form of open source software (see Pattern).

Words such as perfect or prachtig (beautiful) are very positive and should be tagged in green. Green means +1.0. Others such as aanraden (to recommend) are slightly positive. Words such as verschrikkelijk (terrible) or idioot (an idiot) are very negative and should be tagged in red. Red means –1.0. Others such as ingewikkeld (complicated) are slightly negative. Some words such as eigenaardig (peculiar) or hopen (to hope) are subjective without being positive or negative. They should be tagged dark purple. Many words such as wetenschappelijk (scientific) or zijn (to be) are objective, they do not convey a personal opinion. They should be tagged light purple.