Registration Number
0220U001092,
0117U000575
, R & D reports
Title
Fuzzy logic and computational linguistics based the identification of hidden dependencies in online social networks
popup.stage_title
Head
Shtovba Serhiy Dmytrovych,
Registration Date
23-01-2020
Organization
Vinnitsa National Technical University
popup.description2
Object of the research is the information processes in social networks. The goal of the research is creating an information technology for the identification of hidden dependencies in online social networks based on fuzzy logic and computation linguistics the principles. Research methods are system approach, mathematical modeling, fuzzy sets theory, machine learning, computational linguistics. The method for calculation of social network actor profile features is developed. The method produces 2 new features: 1) user response on an automatically created dialogue; 2) a linguistic portrait of the user based on a graph of its texts parsing. A method of classifying social network participants is proposed. The method is based on the frequency of keywords entering the corpus of pre- categorized texts, with the reduction of classes and aggregation of attributes by different keywords. A method of clustering the subject areas of social network information is developed. The method is based on the mapping from vocabulary-free keywords description to basic keywords and creating the lexical ontologies of the text taking into account the analysis of associative relations between its word forms. The new method for keyword selection is proposed. The method is robust to an information noise. The novelity of the method is a correction of meaningful tokens based on syntax of sentences. A method for formalizing the functioning of a social network is proposed. The method is based on the principles of logical-algorithmic description and swarm intelligence. It allows take into account both individual algorithms of the actors' activity and collective decision-making mechanisms in the network. A two-stage fuzzy rule-based method for predicting response of social network users to new information is proposed. The first stage relates to detection the messages with anomalous popularity. The second stage relates to prediction of the number of comments received for an ordinary messages. The results of the research are used in the lecture courses of the Faculty of Computer Systems and Automation of VNTU. Also, the results are used in the development of the SPE "Spilna Sprava", Vinnytsia, for the automatic selection of keywords from text documents.
Product Description
popup.authors
Бісікало Олег Володимирович
Козачко Олексій Миколайович
Кривогубченко Сергій Григорович
Кулаков Павло Ігорович
Лісовенко Анна Ігорівна
Мазуренко Віктор Володимирович
Петричко Микола Володимирович
Ракитянська Ганна Борисівна
Тилець Роман Олексійович
Штовба Олена Валеріївна
Штовба Сергій Дмитрович
Яхимович Олександр Вікторович
popup.nrat_date
2020-04-02
