Sentiment Analysis Reloaded: A Comparative Study On Sentiment Polarity Identification Combining Machine Learning And Subjectivity Features

Title: Sentiment Analysis Reloaded: A Comparative Study On Sentiment Polarity Identification Combining Machine Learning And Subjectivity Features
Authors: Ulli Waltinger
Pub/Conf: Web Information Systems and Technologies – 6th International Conference, WEBIST 2010, Valencia, Spain, April 7-10, 2010

Abstract:
In recent years a variety of approaches in classifying the sentiment polarity of texts have been proposed. While in the majority of approaches the determination of subjectivity or polarity-related term features is at the center, the number of publicly available dictionaries is rather limited. In this paper, we investigate the performance of combining lexical resources with machine learning-based classi er for the task of sentiment classi cation. We systematically analyze four di fferent English and three di erent German polarity dictionaries as a resources for a sentiment-based feature selection. The  evaluation results show that smaller but more controlled dictionaries used for feature selection perform within a SVM-based classi cation setup equally good compared to the biggest available resources.

BibTeX:

@inproceedings{DBLP:conf/webist/Waltinger10,
  author    = {Ulli Waltinger},
  title     = {Sentiment Analysis Reloaded - A Comparative Study on Sentiment
               Polarity Identification Combining Machine Learning and Subjectivity
               Features},
  booktitle = {WEBIST 2010, Proceedings of the 6th International Conference
               on Web Information Systems and Technologies, Volume 1, Valencia,
               Spain, April 7-10, 2010},
  publisher = {INSTICC Press},
  year      = {2010},
  isbn      = {978-989-674-025-2},
  pages     = {203-210},
}

PDFBibTeX