Development of Text Processing Algorithms for Detection of the User Stance in Social Networks

In this work, we have put forward a broad goal of developing a series of efficient text processing algorithms for automatic detection of stance among the social network sites users. Overall, this work makes several theoretical and practical contributions. They can be summarized as follows: — a comprehensive literature review on the topic of sentiment detection was performed, and the new research areas were identified; — the problem of polarity classification was formalized, decomposed into its subtasks, and the requirements towards each of them were formulated; — a novel method for target detection in text messages was proposed, combining the use of a semantic knowledge bases with vector representations of words; — a formal study of polarity scores composition was performed within the framework of SK-languages to model semantics of natural language; — following the determined composition rules, a novel method for sentence-level polarity composition based on dependency structures was proposed; — the method was tested against the current state-of-the art approaches and was shown to be highly efficient; — lastly, they were implemented in the form of a software system, which enables the large-scale analysis of user stance in social network sites. Overall, this work has shown how emergent methods in natural language processing can be used to enhance the performance of the conventional social network monitoring services.