MoSa: A Modeling and Sentiment Analysis System for Mobile Application Big Data
A large amount of data about ending users are generated in the interaction over mobile applications, which becomes a valuable data source for sensing human behaviors and public sentiment trends on some topics. Existing works concentrate on traditional feedback data from web sites, which usually come from desktops instead of from mobile terminals. Few studies have been conducted on interactive data from mobile applications such as news aggregation and recommendation applications. In this paper, we propose a system that can model feedback behaviors of mobile users, and can analyze sentiment trends in mobile feedbacks. The testing data are authentic and are dumped from the most frequently used mobile application in China called Toutiao. We propose several analysis methods on sentiment of comments, and modeling algorithms on feedback behaviors. We build a system called MoSa and by using the system, we discover several implicit behavior models and hidden sentiment trends as follows: During news spreading stage, the number of comments grow linearly per month with slope of 3 in 3 months; The dynamics of replying comments are positively correlated with personal daily routines in 24 h; Replying comment behaviors are much more rare than clicking agreement behaviors in mobile applications; The standard deviation of sentiment values in comments are highly influenced by timing stages. Our system and modeling methods provide empirical results for guiding interaction design in mobile Internet, social networks, and blockchain-based crowdsourcing.
KeywordsMobile big data Sentiment analysis Behavior modeling Mobile applications
The research was financially supported by Major Scientific and Technological Special Project of Guizhou Province under Grant No. 20183001, the Open Funding of Guizhou Provincial Key Laboratory of Public Big Data under Grant No. 2017BDKFJJ006, Open Funding of Hubei Provincial Key Laboratory of Intelligent Geo-Information Processing with under Grant No. KLIGIP2016A05, and National Natural Science Foundation of China under Grant No. 61502362. We also thanks for the comments from W. Jiang, S. Lin, Y. Liao, and M. Lei.
- 1.Delan, X., Juming, C., Shengli, T.: Orientation research based on HowNet. Comput. Eng. Appl. 44(22), 143–145 (2008)Google Scholar
- 3.Shi, H., Jia, D., Miao, P.: Improved information gain text feature selection algorithm based on word frequency information. Comput. Appl. 34(11), 3279–3282 (2014)Google Scholar
- 4.Yang, J., Lin, S.: Emotion analysis on text words and sentences based on SVM. Comput. Appl. Softw. 28(9), 225–228 (2011)Google Scholar
- 5.Fan, N., An, Y., Li, H.: Research on analyzing sentiment of texts based on k-nearest neighbor algorithm. Comput. Eng. Des. 33(3), 1160–1164 (2012)Google Scholar
- 6.Popescu, A.M., Etzioni, O.: Extracting product features and opinions from reviews. In: HLT/EMNLP on Interactive Demonstrations, pp. 32–33 (2005)Google Scholar
- 7.Wei, W., Xiang, Y., Chen, Q.: Survey on Chinese text sentiment analysis. J. Comput. Appl. 31(12), 3321–3323 (2011)Google Scholar
- 9.Yang, X., Ma, Q., Yu, L., Mo, Y., Wu, J., Zhang, Y.: Gauging public opinion with comment-clusters. New Technol. Libr. Inf. Serv. 32(7), 51–59 (2016)Google Scholar
- 10.Yijin, C., Shujin, C., Guihong, C.: Online public opinion mining: user’s sentiment analysis. Doc. Inf. Knowl. 6, 90–96 (2013)Google Scholar
- 11.Zhou, Y., Yang, J., Yang, A.: A method on building chinese sentiment lexicon for text sentiment analysis. J. Shandong Univ. (Eng. Sci.) 6, 27–33 (2013)Google Scholar
- 13.Wang, Z., Wu, Z., Hu, F.: Words sentiment polarity calculation based on HowNet and PMI. Comput. Eng. 38(15), 187–189 (2012)Google Scholar
- 14.Liu, Z., Liu, L.: Empirical study of sentiment classification for Chinese microblog based on machine learning. Comput. Eng. Appl. 48(1), 1–4 (2012)Google Scholar