Abstract
Background. The article is devoted to the issues of effective organization of collection and information analysis about the attitude of Twitter users to brands in the software application form. Issues such as research into modern means of collecting and analyzing information are considered; definition of the functionality that the application should implement; analysis of architectural solutions and selection of software necessary for its implementation. Methods. When conducting research, marketing theory is used in the field of collecting information about consumer opinions, research on methods of information analysis for the purpose of classifying consumer mood, empirical analysis and synthesis of architectures used in the creation and comparison of neural network models for text classification, development and construction of own model for classification. Results. As part of the task of software implementation of tweet text analysis, the architecture of convolutional and recurrent neural networks was investigated, a comparison of various hyper parameter values of neural networks was made, in particular, activation functions, loss functions, the number of learning epochs, the number of network layers, a comparison of different Python libraries for processing natural languages in the context of tweet evaluation. Сonclusions. The practical significance of the study is the creation of a software tool for effective analysis of Twitter users’ attitudes towards brands, which can serve to improve the effectiveness of marketing activities of brands.
Publisher
Taras Shevchenko National University of Kyiv
Reference19 articles.
1. Appel, O., Chiclana, F., Carter, J., & Fujita, H. (2016). A hybrid approach to the sentiment analysis problem at the sentence level. Knowledge-Based Systems,(108). https://doi.org/10.1016/j.knosys.2016.05.040
2. Chandra, P., & Singh, Y. (2004). Feedforward Sigmoidal Networks-Equicontinuity and Fault-Tolerance Properties. IEEE Transactions on Neural Networks, 15(6), 1350-1366. https://doi.org/10.1109/tnn.2004.831198
3. Comparison of static-feedforward and dynamic-feedback neural networks for rainfall-runoff modeling;Chiang;Hydrol,2004
4. Choi, D., Shallue, C. J., Nado, Z., Lee, J., Maddison, C., J., & Dahl, G. E. (2019). On Empirical Comparisons of Optimizers for Deep Learning. Addis Ababa.
5. Conneau, A., Schwenk, H., Barrault, L., & Lecun, Y. (2017). Very Deep Convolutional Networks for Text Classification. Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics, (1), Long Papers, Valencia, Spain. https://doi.org/10.18653/v1/e17-1104