Developing a Machine Learning Model to Search for Self-Organized States of the Twitter Social Network

This dissertation was written as a part of the MSc program Big Data Systems at the Higher School of Economics. Nowadays we faced with exponential growth of daily flood messages of any kind, twenty-thirty years ago we can’t imagine the volume of information we generate per day. The objective scope of this dissertation is to find a way to use the advantages of the semantic meaning of information from these innumerable messages in order to discover what people enjoy/like/share and what don’t. Now we have technological opportunities to understand what they prefer or what they think about some person/situation or event. To achieve this was developed python-based machine learning algorithm which can be used to add sentiment coloring using a system of ratings of part of the sentences. An attempt was made to combine pre-trained on clean data neural network and use it for every data we need. Generally, we can say that aim of this work - to create a pipeline for semantic coloring which will more reliable and robust mechanism. This thesis was completed under the supervision of Senior Lecturer Alexander A. Gorbunov, I would like to thank for his assistance, advice, and support that he has given me throughout the implementation of this thesis. Finally, I would like to thank my family for help and my friends, especially Maria who advised me how to plot this not-easy-to-use dataset and Nikita Zhadaev from Sberbank NLP Team who help me to understand the advantage of using specific neural networks.