How do I create a chatbot using TensorFlow or PyTorch using like the one defined in DialogFlow? What are the best datasets that I can use so to create my own personal assistant like google assistant?
I want to create a chatbot (an open-source project) as an assistant for custom tasks (like google assistant).
I have tried many neural network models like seq2seq but I couldn't get satisfiable results maybe because of the small dataset (I took from Simpsons movie script) or model (seq2seq). I am curious what model they use at google and what type of dataset they pick to give such good results or any normal person can create fully functional chatbots without relying on paid services ( like google's DialogFlow, api.ai, etc.) with good results.
I recently heard of OpenAi's implementation of a specific model named as gpt-2 which as they concluded in the paper showed remarkable performance but they didn't provide the dataset because of some reasons.
What I want to say there are a lot of resources and codes on the internet to make a working chatbot (or maybe that what they show) but when I try to replicate them I always fail to get even remotely close good results.
So I need proper guidance on how to make and train such chatbots with my own laptop (with 16GB RAM, 2GB GPU . I can even get better configuration) and no other money spent on google services or any such paid API's.
Please suggest something if someone got good results.