Automating Event Extraction: A Guide to Identifying Triggers, Arguments, and Roles

In this paper, we present CooperKGC, a novel knowledge graph construction (KGC) approach that leverages the power of language models (LLMs) to enhance the efficiency and accuracy of information exchange in collaborative settings. Our proposed method combines three key components: a message simplification module, an event extraction agent, and a relation extraction agent. These agents work together to streamline the KGC process by identifying relevant events and relationships within unstructured natural language texts.
The message simplification module uses LLMs to extract the most critical information from input texts, while the event extraction agent relies on a combination of trigger words and argument roles to identify potential events in the text. The relation extraction agent then refines these events by linking them to relevant entities and attributes, ensuring that the resulting knowledge graph is accurate and comprehensive.
We evaluate CooperKGC through extensive experiments using various benchmark datasets, demonstrating its superiority over existing KGC methods. Our results show that CooperKGC can effectively identify and extract complex events and relationships, while also improving the overall efficiency of the KGC process. With its ability to simplify knowledge graph construction, we believe that CooperKGC has significant potential for real-world applications in fields such as natural language processing, information retrieval, and data integration.