Article Preview
Top1. Introduction
Microblogging service helps users send and read multiple realtime messages or news feeds getting popular as it connects with the globalized world, which we know as Twitter. The news feed composes a realtime instantaneous communication medium for everyday users. Twitter sends out more than 700 million tweets by more than 400 million everyday users. Having such a connected world tool in users' hand, event tracking such as a football game creates interest among the new generation. Consequently, users follow real-time news, business growth proposals, and stock and crypto market updates as social events. The emergent populous nature of the social world shows another interest related to emergency event tracking such as disasters, disease spread, and terrorist attacks. Users tweeting information regarding events on social media can generate a lot of perspective in events compared with the news media. However, a plethora of events and anomalies are reported every hour on Twitter, and news portals miss the majority of personal users news. for example, family members update messages of their health about wellness or care for their loved ones.
Users' textual perceptions of their surrounding environmental conditions or emotions are people-centric sensing data and are often interested in detecting a sequence of crucial moments. This people-centric data sensing from social networks evolve for an event that spans time. However, due to the rate of data generation in the social world, analyzing and summarizing data topics for a specific event and sub-event anomalies is a challenge. This challenge is due to noisy data or content in heterogeneous social networks. This problem is widely studied (Srivastava et al., 2012). However, the preciseness to detect all the events and subevents is low besides identifying essential moments. This shows that the fundamental requirement of social data analysis is to address unique requirements such as duration of tweets, emotions, and geographic location (Srivastava & Sankar, 2020), which are dynamic in nature. Therefore, this makes a clear path for the summarization tasks consisting two parts: (1) Detecting a stream of subevent anomalies in an event. (2) A generating module can categorize the events and provide a summary for subevents descriptions. In this paper we propose a novel self-sufficient system that deals with the challenges mentioned above in social media event detections. Our system categorizes the data into actions, emotions, and locations and decomposes these events into time spaced graph since we assume that with time there will be the change in categories size and details (e.g., users use the same tweet to add more details). This decomposition are meta tasked through a common source, whether the users follow or are added as a friend (Le Wang et al., 2012). We will also provide an overview of metrics on improving the credibility of social sensing. This social sensing can indeed provide better opportune characteristics in finding false negative rumors.
The rest of this paper is organized as follows. Section II describes earlier work done in ðeld of source selection and fact-ðnding. Section III presents the anomaly detection and categorization problem and proposes a set of source selection schemes that diversify the sources admitted for data collection purposes. Section IV provides the problem statement for subevent detection and possible solution. Finally, evaluation results demonstrating the effect of source selection on credibility assessment of collected data is presented in Section V followed by conclusions in Section VI.