Mehdi Azarafza - Department of Computer Engineering, Faculty of Electrical & Computer Engineering, University of Tabriz, Tabriz, Iran,
Mohammad-Reza Feizi-Derakhshi - Department of Computer Engineering, Faculty of Electrical & Computer Engineering, University of Tabriz, Tabriz, Iran
Moosa Bagheri Shendi - Department of Computer Engineering, Faculty of Electrical & Computer Engineering, University of Tabriz, Tabriz, Iran

Bursty events and keyword extractions from microblogs are a sensitive activity for identifying and extracting top words which are descriptive candidates from text. Bursty event detection is mainly used to express hot event quickly which is discussed in texts and understanding the main orientation of the texts/writers from social media. Each various keyword extraction algorithms prepared for several languages as English, France, and Russian, but for the Persian language this issue is still alive. This presented work is focused on a proposed method for keyword extractions of Persian language that are used for data extractions from Persian microblogs. To evaluate the precision and proposed algorithm, 4 different cases are applied to extract the keyword as descriptive candidate of texts. Due to the lack of reliable measure for Persian language, proposed method is justified by expert system, TF-IDF, Gensim and latent dirichlet allocation (LDA) methodologies. The experimental results indicated that proposed method extracted approximately 60% of top Persian keywords which rates considerably more than the others.

Text-Rank, Persian Keyword Extraction, Microblogging, Data stream.