I have created an ML pipeline where we could feed in any unstructured datasets and through Named Entity Linking and relationship extraction, I can generate a graph using those results in Neo4j. I am trying to evaluate the graph generated, but I am quite unsure of how to proceed further with this. Is there any way of evaluating an automatically generated graph through unstructured datasets without any ground truth?
How to find the Accuracy and F1 score of the graph?
I was trying to create something similar to the one below blog
https://towardsdatascience.com/from-text-to-knowledge-the-information-extraction-pipeline-b65e7e3027...
I would like to know more about evaluating the performance of the same.
