This repo contains various methods explored for Conversational Q&A System on CoQA Dataset
- FlowQA: Implemented LSTM and GRU-based Flow-QA architecture
- GraphFlow: a GNN based architecture with various embeddings
- Transformer: BERT base and BERT large architectures
In total we performed 13 different experiments and found Transformer based architecture gives the best F1 score on the validation set.
The Presentation explaining the methods and results can be found here
Note: To run individual methods check the ReadMe for each methods and install requiremenets from requirement.txt files. Python version used for all the experiments: Python 3.8