Analysis of tweets to find the basis of popularity based on events semantic similarity

Purpose The purpose of this paper is to propose and experiment a framework for analysing the tweets to find the basis of popularity of a person and extract the reasons supporting the popularity. Although the problem of analysing tweets to detect popular events and trends has recently attracted extensive research efforts, not much emphasis has been given to find out the reasons behind the popularity of a person based on tweets. Design/methodology/approach In this paper, the authors introduce a framework to find out the reasons behind the popularity of a person based on the analysis of events and the evaluation of a Web-based semantic set similarity measure applied to tweets. The methodology uses the semantic similarity measure to group similar tweets in events. Although the tweets cannot contain identical hashtags, they can refer to a unique topic with equivalent or related terminology. A special data structure maintains event information, related keywords and statistics to extract the reasons supporting popularity. Findings An implementation of the algorithms has been experimented on a data set of 218,490 tweets from five different countries for popularity detection and reasons extraction. The experimental results are quite encouraging and consistent in determining the reasons behind popularity. The use of Web-based semantic similarity measure is based on statistics extracted from search engines, it allows to dynamically adapt the similarity values to the variation on the correlation of words depending on current social trends. Originality/value To the best of the authors’ knowledge, the proposed method for finding the reason of popularity in short messages is original. The semantic set similarity presented in the paper is an original asymmetric variant of a similarity scheme developed in the context of semantic image recognition.