Graph Convulational Network (GCN) is a technique applied in several studies to analyze events organized in a graph structure. Using this technique as a basis, we created an algorithm that adds a new parameter to its analysis, with the power to increase or decrease the importance of certain types of data, depending on their classification among components of an event (why, what, , when, where and how) thus altering the final result of the experiment. Note in the image below, after using the importance P Matrix, the relevance of the 'where' parameter data in obtaining the final result decreased.
Our model was treined with To use the pre-traned model to analise event embeddings we use datasets from different domains.
This is a algorythem under development. If you fide any error or unknown problems in the code, we would be glade do have your help do make it easier to use.