AmexBot from scratch with numpy

A Deep Neural Network based chat bot built from scratch using numpy

Test the chatbot

Run the chatbot in command line : python3 bot.py Type "quit" to quit the chatbot. Type any questions about one of the 7 topics :

• Greeting
• Historical
• Food
• Animals
• Cruise
• Night tour
• Relaxation
• Show concerts

To get inspiration look at the csv file "Test.csv". You can start by asking questions from the file. The neural network hasn't been trained on a very exhaustive dataset, some unrelated questions or unknown words could be hard to understand for the chatbot.

Look at the code

A raw python file coded in an object-oriented style can be found in "scratch.py".

A more comprehensive notebook can be found in "scratch_explained.ipynb" that explains in more detail the architecture of the deep neural network.

Training process

Look at the results of the training in "train.ipynb"

Training process complexity. Graph of training time as a function of the training size.

User load performance

Graph of prediction time as a function of the testing size.