An Efficient Model for Mining Outlier Opinions

In the internet era, opinion mining became a critical technique used in many applications. The internet offers a featured chance for users to express and share their views and experiences anywhere and at any time through various methods as online reviews, personal blogs, Facebook, Twitter and companies’ websites. Such treasure of online data generated by users play an essential role in decision-making process and have the ability to make radical changes in several fields. Although the opinionated text can provide significantly invaluable information for the wide community either are individuals, business, or government, the outlier or anomaly opinions could have the same impact but in opposite manner which harm these fields. Consequently, there is an urge to develop techniques to detect the outlier opinions and avoid their negative impacts on several application domains which rely on opinion mining. In this paper, an efficient model for mining outlier opinions has been proposed. The proposed MOoM model, stands for Mining Outlier Opinion Model, offers for the first time the ability to mine outlier opinions from product’s free-text reviews. Accordingly, it can help the decision makers to improve the overall sentiment analysis process and perform further analysis on the outlier opinions to get better understanding for them and avoid their negative impact. The proposed model consists of three modules; Data preprocessing module, Opinion mining module, and outlier opinions detection module. The proposed model utilizes the lexicon-based approach to extract sentiment polarity from each review in the dataset. Also, it uses the Distance-based outlier detection algorithm to produce a graded list of review holders with outlier opinions. Experimental study is presented to evaluate the proposed model and the results proved the model’s ability to detect outlier opinions in the product reviews effectively. The model is adaptable to be used in other fields rather than product’s reviews by customizing its modules’ layers.

@article{Hassan2020,
  title     = {An Efficient Model for Mining Outlier Opinions},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {11},
  number    = {5},
  year      = {2020},
  publisher = {The Science and Information Organization},
  author    = {Neama Hassan and Laila A. Abd-Elmegid and Yehia K. Helmy},
  doi       = {10.14569/IJACSA.2020.0110522},
  url       = {https://doi.org/10.14569/IJACSA.2020.0110522}
}