A comparative study of different sentiment lexica for sentiment analysis of tweets
|Conference paper (help)|
|A comparative study of different sentiment lexica for sentiment analysis of tweets|
|Authors:||Canberk Özdemir, Sabine Bergler|
|Citation:||Proceedings of Recent Advances in Natural Language Processing : 488-496. 2015 September|
|Web:||DuckDuckGo Bing Google Yahoo! — Google PDF|
|Article:||Google Scholar PubMed|
|Restricted:||DTU Digital Library|
A comparative study of different sentiment lexica for sentiment analysis of tweets is an evaluation of sentiment analysis systems using the NRC lexixon (EmoLex), Bing Liu's lexicon, MPQA and AFINN (written "aFinn" in the paper) as well as their own lexicon Gexi. They "add negation and modality sensitive features", and applies it on SemEval 2015 Task 10B and Task 11. In task 11 they performed well as one out of 35.
They use a machine learning method for predicting sentiment applying libSVM with an RBF kernel. and a combination of 12 sets of features derived from sentiment word lists POS-tagging and others, e.g., for AFINN they use positive score, positive negated, positive modality score, etc ) page 493)
- SemEval 2015 Task 11: 0.768 as best with a combination of features and machine learning.
- "the smallest lexicon, aFinn, is the best solo performer."