Text Classification Improvement using Knowledge Graphs
This project focusses on improving the performance of BERT using Knowledge Graphs
The main files are
BERT_NoKGE: This file consists of code to test the baseline model performance for all the datasets considered. It also can be used to test the datasets with the String Manipulation Approach BERT_KGE: This file consists of code to test the BERT model with Knowledge Graph Embeddings(KGE) It consists of 3 models which can be used to improve the performance of BERT
The supporting files are
datasets/ : This folder consists of all the datasets used to test the models created. kg-embeddings/ : This folder consists of the pickle files used for embeddings data-prep/ : This folder consists of the files used for data preparation as needed by the main files