SteliosKliafas / Finetuning-Transformer-Models-Movie-Plot-Generation-Classification

Used the Netflix Movie & TV Shows dataset to finetune the GPT-2 model for Text Generation, while also finetuning the BERT model for classification to evaluate and compare the model’s accuracy on the GPT-2 generated text and the original

INM706 Coursework - Deep Learning for Sequence Analysis

Sarah Rhalem (190051884) & Stelios Kliafas (200041011)

Finetuning Transformer Models for Movie Plot Generation & Classification

Main notebook: INM706_Code_SR_SK.ipynb

All packages and the versions required to run are detailed in the main notebook. Specifically the following packages are required: --pip install transformers --pip install bertviz

Model checkpoint one-drive location: https://cityuni-my.sharepoint.com/:f:/g/personal/sarah_rhalem_city_ac_uk/Ekc0pwywzSlKucWxQr4Br-QB_j3NeZJ9JPm5CsKKDO1TpA?e=OeOx6z

All modules are contained with the modules folder. Results are stored as pickle files in the Results folder. Figures used for the report are stored in the Figures folder.

Datasets are contained within the Data folder:

• Original dataset Netflix_titles.csv or can be located at https://www.kaggle.com/shivamb/netflix-shows?select=netflix_titles.csv
• Processed data file: netflix_plot_dataset.txt
• GPT-2 generated plot datset