NLP-domain-specific-assistant

NLP based domain specific assistant

A NLP based ML assistant.

Details of project:

• Tensorflow 1.5
• Bidirectional encoder
• Attention
• Beam search decoder for Inference
• Data set - Cornell movie corpus

Setup/Requirements:

• Python 3 - brew install python3
• Tensforflow 1.5 installation using - https://www.tensorflow.org/install/
• Activating Tensorflow virtual env - $ cd targetDirectory $ source ./bin/activate
• Numpy

To Execute the project:

• Training: python3 main.py training

• Testing: python3 main.py testing - used for the automatic evaluation of the responses using BLEU

• Prediction:

  • Input file mode - python3 main.py inference file OR python3 main.py inference
  • Command line mode - python3 main.py inference command

• Data preparation (if needed):

  • movie_conversation.txt and movie_lines.txt should be present under /Data
  • Run python3 data_preparation.py

To run TensorBoard:

• Activate tensorflow virtual environment
• tensorboard --logdir=locationOfmodelDirectory

Add a directory for model:

• /model/seq2seq to store checkpoints