This paper presents a hybrid approach to sentiment classification method for Korean texts. It is based on a cascading system by which lexicon-based classification first conducts the sentiment detection along with the local parsing of sentiment constituents, and a supervised machine learning algorithm sorts the texts out of the lexicon. We use a fine-grained Korean machine-readable dictionary for the lexicon-based classification, dealing with Polarity Shifting Devices (PSDs) which are divided into Intensifier, Switcher, Activator, and Nullifier. By structuring PSDs and polarity values of opinion texts, it is possible to process complex sentiment constituents efficiently, such as a structure resulting from double negation. Through the performance evaluation, we prove this hybrid approach particularly enhanced by sentiment lexicons and PSDs outperforms the baselines.
@InProceedings{YOO18.6, author = {Gwanghoon Yoo and Jeesun Nam}, title = {A Hybrid Approach to Sentiment Analysis Enhanced by Sentiment Lexicons and Polarity Shifting Devices}, booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)}, year = {2018}, month = {may}, date = {7-12}, location = {Miyazaki, Japan}, editor = {Kiyoaki Shirai}, publisher = {European Language Resources Association (ELRA)}, address = {Paris, France}, isbn = {979-10-95546-24-5}, language = {english} }
Summary of the paper