Classification of the cyber texts and comments into two categories of positive and negative sentiment among social media users is of high importance in the research are related to text mining. In this research, we applied supervised... more
Classification of the cyber texts and comments into two categories of positive and negative sentiment among social media users is of high importance in the research are related to text mining. In this research, we applied supervised classification methods to classify Persian texts based on sentiment in cyber space. The result of this research is in a form of a system that can decide whether a comment which is published in cyber space such as social networks is considered positive or negative. The comments that are published in Persian movie and movie review websites from 1392 to 1395 are considered as the data set for this research. A part of these data are considered as training and others are considered as testing data. Prior to implementing the algorithms, pre-processing activities such as tokenizing, removing stop words, and n-germs process were applied on the texts. Naïve Bayes, Neural Networks and support vector machine were used for text classification in this study. Out of sample tests showed that there is no evidence indicating that the accuracy of SVM approach is statistically higher than Naïve Bayes or that the accuracy of Naïve Bayes is not statistically higher than NN approach. However, the researchers can conclude that the accuracy of the classification using SVM approach is statistically higher than the accuracy of NN approach in 5% confidence level.
This paper presents an application of text mining to support Competitive Intelligence (CI). A case study was built using the ETO System, which enables trade-related bodies to exchange information by e-mail. As this information is... more
This paper presents an application of text mining to support Competitive Intelligence (CI). A case study was built using the ETO System, which enables trade-related bodies to exchange information by e-mail. As this information is available in textual formats, we used text-mining tools to support the CI process. The strategy uses domain knowledge to extract concepts from the texts. A mining tool searches for patterns in concept distributions and correlations, aiding the identification of strategic information. The main contribution of the paper is the use of an inexpensive strategy to allow a competitive advantage to Small and Medium Enterprises with minimal cost.