search

Conchylicultor / DeepQA

My tensorflow implementation of "A neural conversational model", a Deep learning based chatbot
Apache License 2.0
2.93k stars 1.17k forks source link

I got an idea..... #138

Open SeekPoint opened 7 years ago

SeekPoint commented 7 years ago

for a large dataset about 10M QA pairs would it be a better performance on accuracy if we divide the dataset by the length of sentences. and feed it to different training model and decoding it accordingly(maybe different parameters on RNN size, layers for the different model) ?

any comments!!!!

Conchylicultor commented 7 years ago

The recent versions of TensorFlow implement a system of bucket which will batch together sentences of similar lengths. This is done mainly for performance purpose. The model stays the same for any sentence length.