Construction of a News Graph and Services Around it

This paper presents a method of constructing a causal relationship graph based on news texts on the topic "Economy" for the year 2023. The main focus is on finding causal relationships between events presented in Russian-language news. Modern text summarisation and analysis models including FRED-T5 and gpt-3.5 were used for data extraction. After data collection and processing, cluster analysis was performed using Agglomerative Clustering model to reduce the number of nodes in the graph. The final graph contains 7,427 nodes and 36,145 edges, allowing for effective visualisation and analysis of key economic events and their interrelationships. The paper also explores different ways of using the resulting graph such as direct visualisation, temporal dynamics analysis and the application of Retrieval-Augmented Generation (RAG) to improve query processing. This approach demonstrates a new methodology for the extraction and visualisation of causal relationships relevant for the analysis of Russian-language texts.