FlixREC: A Movie Recommendation Chatbot

Created by: Thanoshi Balasuriya, Christa Mumley, Smita Rosemary, and Gauri Sarin

Northeastern University | CS4100: Artificial Intelligence

---

Setup

Without retraining the chatbot model:

1. Make sure cs4100-movie-recommendation-engine is downloaded either through git clone or just downloaded.
2. Make sure tkinter is installed it normally comes installed as a standard but just in case run brew install python-tk in the terminal.
3. Go to this Google Drive folder
4. Download the zipped "data" file. Unzip into cs4100-movie-recommendation-engine/chatbot/data
5. Download the zipped "movie_model_data" file. Unzip into cs4100-movie-recommendation-engine/Models/saved_weights

With training the chatbot model:

Open terminal > cd to \chatbot > run TrainChatbot.py

Note: re-training may not work if you don't have pytorch cuda enabled.

This trains question intents from intents.json. If you want to add tags/fine-tune the intents, change that file.

Running the program:

Open a terminal window in the root folder (/cs4100-movie-recommendation-engine)

Run: python main.py

---

File Organization:

Chatbot - files related to training the Chatbot, using BERT

Datasets - .csv files of data

GUI - files related to the graphical UI

Models - files related to the actual recommendation models. Models/Jupyter Notebooks is a folder of the original notebooks used to create the models

controller.py - handles the input and directs program to right action/recommendation model

main.py - runs the Chatbot from the terminal

---

General Notes

• Loading the program may take a little while
• Some file paths are currently in MacOS format: for Windows, in GUI/chat_gui.py, replace / with \\ in the file paths
• Jupyter Notebooks (seen in Models/Jupyter Notebooks) are where we brainstormed/fleshed out our ideas - the file paths within are not currently accurate (you will run into errors attempting to run the notebooks)