Age Restriction Analysis of Movie Scripts

The goal of our project is to automate and extend the age rating of movies based on the script of the movie. The criteria for our evaluation will be primarily based on the official rating rules used by the Motion Picture Association.

1. General information

Team Members:

• Jakob Hennighausen
  • Email: jakob.hennighausen@stud.uni-heidelberg.de
• Davit Melkonyan
  • Email: davit.melkonyan@stud.uni-heidelberg.de
• Leon Remke
  • Email: leon.remke@stud.uni-heidelberg.de

    Existing code fragments

Utilized libraries

• Elastic search
• React
• Fastapi
• Spacy
• Sklearn Check requirements.txt in sub-modules of project for detailed information.

Contributions

See "Project log" section.

2. Project State

Planning State

One of our high level milestones for November could not be achivet yet. Find a detailed list of goals achived and not achived.

Achived Goals

• Basic project setup
• Setup elastic search hosting on linux server
• Data crawling
• Data understanding
• Data analysis
• Start of data preprocessing
• Get baseline for PG-Ratings

  Open Goals

• Data preprocessing pipeline
• Decision which statistical method to use for age ratings
• First PG ratings on films

  Future Planning

Currently, the project is behind schedule with respect to the initial milestone plan. This should be made up by the lecture break from 22.12.22 to 07.01.23. One reason for the delay is the change in data sourcing. The operator of a platform for film scripts had unexpectedly stopped responding.

High-level Architecture Description

High level application architecture

High level processing architecture

todo: add descriptions to preprocessing steps and knowledge from data analysis

Experiments

First experiment to find the official age ratings of movies through a TMDB API. For this we wrote a python script and used a selection of movie titles to get the age ratings. Results can be found here

Data analysis, explocation and description. Results can be found in next section.

3. Data Analysis

This section can be found in this jupyter-notebook:

Project log

Davit

• POC dataset crawling
• Linux server for elasticsearch instance
• Data understanding, exploration and preprocessing
• Labels scraping
• Model fine tuning
• Random Forest Implementation

  Jakob

• Research papers and materials
• Research for baseline dataset
• Implementation of finding age ratings by movie titles with TMDB API
• Implementation of RNN based on paper
• Preprocessing
• Creation of training and test datasets

  Leon

• Setup git project (react, fastapi, elasticsearch, dockerfiles)
• Setup git-hooks integration
• Run elasticsearch instance on linux server
• Frontend development
• Fastapi development
• SVM Classifier
• Elasticsearch setup and upload

  Project structure

  Find a raw overview how the project is structured:

  assets

  Documentation files

  data

  Training data and model results

data_exploration

Data exploration notebook

data_gathering

• Notebooks for scraping and collecting relevant data (e.g. age ratings)
• Elasticsearch upload script

data_preprocessing

• Notebooks to preprocess data

elasticsearch

• Nginx config
• Documentation
• Elasticsearch was set up on our server. If you want to access the elasticsearch instance, please contact us.

fastapi

• API project

own_model

• SVM and Random Forest model implementation

react_ageflix

• Frontend project

severity_model

• Model from referenced paper

  Presentation

  Presentation can be found in the assets folder.