Intelligent Chatbot using Deep Neural Network

Domain : Artificial Intelligence, Machine Learning, Natural Language Processing

Sub-Domain : Deep Learning, Conversational Agent, Language Modeling

Techniques : Sequence-to-Sequence (Seq2Seq), Recurrent Neural Network, Bidirectional RNN, LSTM, Neural Attention Mechanism, Beam Search, Neural Machine Translation, TensorFlow

Description

1. Developed chatbot using encoder and decoder based Sequence-to-Sequence (Seq2Seq) model from Google’s Neural Machine Translation (NMT) module and Cornell Movie Subtitle Corpus.
2. Seq2Seq architecture built on Recurrent Neural Network and was optimized with bidirectional LSTM cells.
3. Enhanced chatbot performance by applying Neural Attention Mechanism and Beam Search.
4. Attained testing perplexity of 46.82 and Bleu 10.6.
5. Developed backend using Python and front-end using Python and PyQT.

Intelligent Chatbot Graphical Interface:

Sample Conversations:

Languages : Python

Tools/IDE : Anaconda

Libraries : Neural Machine Translation, TensorFlow, PyQT

Duration : Jan – May 18

Current Version : v1.0.0.0

Last Update : 05.01.2018