How does one add your data augmentation queries to the pipeline? do you just create a training json file full of the fake queries and then feed it to the preprocessor and train along with the real queries? The paper is unclear about what configuration achieved your best result.
How does one add your data augmentation queries to the pipeline? do you just create a training json file full of the fake queries and then feed it to the preprocessor and train along with the real queries? The paper is unclear about what configuration achieved your best result.