The detection and analyzing the events from streams of social media text is very interesting and fascinating research work. The social media platforms become the global platforms for information sharing. These information can be sensed for event bursts detection. After detection of the event further analysis is to be done for extraction of temporal and spatial details of the event. Information extraction from the text is primarily a natural language processing (NLP) task. The emergence of various pre-trained language models (PLMs) in the NLP arena makes NLP solutions efficient. However, despite these improvements, PLMs show a concerning propensity to cause hallucinations, leading to results that are not aligned with input or actual facts. This work focuses on identifying events along with time and location information from social media text with the help of a PLM model, i.e., BERT along with mitigation of hallucination. The experimental results show that the proposed framework improvises a recall metric which can be considered as a hallucination indicator over traditional approach. The recall metric is used as indication because it measures how truly a model can identify positive instances. The better the recall value, the less.