Closed dimwael closed 4 years ago
You can do it in the same way as with other ranking models. Please, see the documentation https://deeppavlov.readthedocs.io/en/master/features/models/neural_ranking.html#training-and-inference-on-your-own-data
. That is to train on your own csv file you need to provide train, dev, test files in the appropriate file format and use the siamese_reader
component as a reader and the siamese_iterator
as an iterator in your configuration file.
Please correct me if I am wrong, instead TF_IDF we can use Bert for classification like for the insuranceQA V1 ? (configs.rankings.ranking_insurance_interact model ) If yes, how can we train it on csv file ?