I was wondering if the
T5 models can do the task of generating sentences in English. Most of the models I have mentioned are trained to translate from English to German or French. Is it possible that I can use the output of
BERT as an input to my Decoder? My theory is that when I already have the trained
Embeddings, I do not need to train the Encoder part. I can just add the outputs of sentences to the decoder to generate the sentences.
In place of finding the loss value from the translated version, Can I compute loss on the reply of a sentence?
Can someone point me toward a tutorial where I can use the
BERT output for the decoder part? I have a data of conversation with me. I want to build a
Chatbot from that data.
I have already implemented
Sequence2sequence model but it is not providing me satisfactory answer.
After some research, 2 such models are there as
BART which are based on the same idea.
If possible, can someone tell me how can I use
T5 to make a conversational bot?