Online political discussions have received lot of attention over the past years. In paper we compare two sentiment lexicon to classify the sentiment of from political discussions. The approach is based on applying the of words between the target and sentiment words to weight the sentence score. The second approach based on using the shortest paths target and sentiment words in a graph and linguistically motivated patterns expressed as dependency. The methods are tested a corpus of sentences from online Norwegian discussions. The results that the method based on dependency performs significantly better the word-based approach.
No takes yet. Share an insight, caveat, or question.
Hammer et al. (2014) studied this question.