Thematic Event Clustering and Storyline Analysis Based on Multi-source Heterogeneous Chinese International Negotiation Documents

In the international relations and foreign policy research, international negotiation data is of great importance for assisting scholars in analyzing patterns of interaction between nations, conflicts of interest, and prospects for cooperation. The analysis of events and their storylines within international negotiation data is instrumental for better managing international relations, formulating relevant policies, and predicting economic trends, which holds substantial importance. Current research surrounding event comprehension is primarily geared towards general news or financial domains, focusing mainly on single-document studies, which do not effectively cater to the study of multi-source heterogeneous international negotiation documents as encountered in reality. Therefore, this work has involved the collection of relevant negotiation documents in the economic and agricultural domains. By the combination of key techniques including clustering algorithms, in-context learning and time series analysis, we completed the construction of event storylines from single to multiple documents in international negotiation data. Based on these event storylines, we also analyzed the relevant key issues and main trends, deepening the understanding of document information. We validated the effectiveness of our workflow on our collected data. Furthermore, we developed a system based on the proposed workflow and implemented visualization, making the analysis of international negotiation documents more convenient and user-friendly.