The purpose of this paper is to study the application of causality in matter-of-fact mapping, and improve the inadequacy of traditional mapping that relies only on event co-occurrence relations. By introducing the causal inference method, we construct a more accurate causal matter-of-fact mapping and validate it in practical applications. The experimental results show that the deep learning model based on BERT (Bidirectional Encoder Representations from Transformers) achieves an F1 value of 0.875 in causal extraction, which outperforms CNN (Convolutional Neural Network)'s 0.825 and RNN's 0.825. In terms of graph optimization, PageRank optimization increases node importance from 0.05 to 0.075, HITS (Hyperlink-Induced Topic Search) optimization increases graph connectivity from 0.70 to 0.85, and the average path length is shortened from 3.5 to 3.0. It can be seen from the data conclusions that optimized causal graph is more effective and precise in representing event relationships and logical structures.