Methodologies of Internet portals users' short messages texts authorship identification based on the methods of mathematical linguistics

摘要

The article deals with the peculiarities of Internet portals, blogs and websites short messages texts authorship determination. The article focuses on possibility to search people who have several different accounts and send messages from them. Sentences dependence on the number of words in portals users' comments is represented. The model of Internet portal text message is provided. Method of Internet portals users' short messages texts authorship identification based on the naive Bayesian classifier is represented. The specific feature of the proposed method is not only frequency dictionary analysis based on messages selection to identify users, but their usage of rules and connections on the base of language syntactic information. The parts of speech frequency and connection frequency between parts of speech are given. The communication graph of parts of speech connections of limited natural language in commentaries is represented. Linguistic characteristics used to identify portal user are given. Structures are distinguished on the base of the communication graph between parts of speech as regards noun prepositional casal form of limited natural language used to identify text authorship. The experiment showing achievable indicators of Internet portal user identification probability depending on training sample is carried out. Probability diagrams of authorship identification based on selected characteristics are represented.