GEOGRAPHIC LOCATION PREDICTION IN BLOGS

Morozov E.V, Bogdanova D.N

208 KB

This paper presents research on geographical lexical variation detection. We present a new corpus of Russian blogs labeled with geographical information. The corpus was extracted from LiveJournal (http://livejournal.com). Only those blogs that contained enough information about the author were used in the experiments. We have performed outlier detection, and thus, removed spam and other irrelevant data. We have studied various feature sets and performed classification based on Support Vector Machine and Naïve Bayes algorithms. The obtained results show that geographic location prediction is a hard task, and many of the blogs do not contain enough information to determine location of their authors, even though in certain cases accurate classification is possible.

Библиографическая ссылка

Морозов Е.В, Богданова Д.Н ОПЕРЕДЕЛЕНИЕ РЕГИОНА АВТОРА ПО ДАННЫМ ЖИВОГО ЖУРНАЛА // Научное обозрение. Физико-математические науки . 2020. № 1. С. 43-43;
URL: https://physics-mathematics.ru/en/article/view?id=55 (дата обращения: 24.06.2026).