Comparative analysis of the results obtained when solving the problem of analyzing the sentiment of a text using convolutional and recurrent neural networks

In the modern world, there are various means of communication: electronic devices, web and mobile applications (Internet forums, chats, blogs, social networks). As a result, a huge amount of information appears about the users themselves, about their attitude to other people, to events taking place in the world. This information can be useful in modeling the processes occurring in society, predicting the behavior of people. Thus, the methods of collecting and analyzing information contained on the Internet are interesting for research. Information on the Internet is presented in the form of a text in a natural language, therefore it is necessary to use the methods of computational linguistics. For example, let's say we have text. Without reading, is it possible to understand what emotion he carries? You can, for example, classify an emotion into positive and negative. The paper discusses Convolutional Neural Networks, which were originally developed for image processing, but also cope with tasks in the field of automatic word processing and Recurrent Neural Networks, the main difference from traditional ones is the logic of the network operation, in which each neuron interacts with itself.

1. Rubtsova Yu. Automatic construction and analysis of the corpus of short texts (microblogging posts) for the task of developing and training a tone classifier. Knowledge Engineering and Semantic Web Technologies. 2012;1(2):109-116.

2. Semina T.A. Sentiment analysis of the text: modern approaches and existing problems. Applied linguistics, translation, translation studies. 2020;2(2):47-64.

3. Sikorsky O.S. An overview of convolutional neural networks for the image classification problem. New information technologies in automated systems. 2017;17(4):17-26.

4. Rubtsova Yu.V. Building a text corpus to customize a tone classifier. Software products and systems. 2015;1(109):72-78.

5. Benderskaya E.N., Nikitin K.V. Recurrent neural network as a dynamic system and approaches to its training. Computer science, telecommunications and management. 2013;3(3):29-40.

6. Nemaltsev A.S. Using recurrent neural networks to analyze raw multilingual text. International Journal of Humanities and Natural Sciences. 2020;3(2):55-59.

7. Cliche M. Twitter Sentiment Analysis with CNNs and LSTMs. Proceedings of the 11th International Workshop on Semantic Evaluation. 2017;11(24):573-580.

8. Mikolov T. Distributed Representations of Words and Phrases and Their Compositionality. Advances in Neural Information Processing Systems. 2013;5(11):3111-3119.

9. Rosenthal S., Farra N., Nakov P. Sentiment Analysis on Twitter. Proceedings of the 11th International Workshop on Semantic Evaluation. 2017;11(24):502-518.

10. Zhang Y., Wallace B. Sensitivity Analysis of (and Practitioners' Guide to) Convolutional Neural Networks for Sentence Classification. arXiv preprint arXiv: 1510.03820. 2015.