We propose the Rule-Based Emission Model (RBEM) algorithm for polarity detection. RBEM uses several kinds of heuristic rules to create an emissive model on polarity patterns. We extensively experiment with our approach on English and Dutch messages extracted from Twitter. Thus we also illustrate that RBEM can be used in multilingual settings and is applicable to social media characterized by use of not always regular language constructs. We demonstrate that designing such an algorithm instead of applying the state-of-the art general purpose classification techniques is a reasonable choice for the automated sentiment classification in practice. Using RBEM we can design a competitive multilingual sentiment classification system showing promising accuracy results of 78.8% on the considered datasets. We provide some further evidence that RBEM-based systems are easy to debug, improve over time and adapt to new application domains.
|Title of host publication||Proceedings of the Second International Workshop on Issues of Sentiment Discovery and Opinion Mining (WISDOM'13, Chicago IL, USA, August 11, 2013; in conjunction with SIGKDD'13)|
|Place of Publication||New York NY|
|Publisher||Association for Computing Machinery, Inc|
|Publication status||Published - 2013|