Event extraction from textual documents is a crucial task in various fields, including criminal investigations, healthcare, and news analysis. In this study, we compare two methodologies for extracting and representing events: a traditional Natural Language Processing (NLP) pipeline and Large Language Models (LLMs). The traditional approach leverages tools such as Flair and SpaCy for Part of Speech tagging, Named Entity Recognition, and syntactic parsing, while the LLM based method utilizes fine tuned models to extract event related entities and resolve coreferences. We construct event graphs to analyze relationships between extracted events, enabling structured representation of temporal and contextual connections. Experimental results on diverse datasets including news articles, literary texts, and multilingual corpora highlight the strengths and limitations of both approaches. While the traditional method excels in structured event extraction with well defined rules, LLMs demonstrate superior adaptability in handling implicit subjects and entity disambiguation. Furthermore, we explore contradiction detection within event graphs to identify conflicting information across multiple sources. Our findings suggest that a hybrid pipeline, integrating rule based NLP techniques with LLM refinement, could enhance event extraction accuracy and contextual understanding.