This is a fine-tuned T5-small model for abstractive summarization. This model is fine-tuned with XSUM dataset and CNN/DailyMail dataset where only 20% of the train split and test split are used due to the limited resources for fine-tuning the model (the fine-tuning is done via Google Colab freeplan). The way the datasets are used is by concatenating the the train split and the test split of each datasets. This model achieved 0.25 in ROUGE-1, 0.08 in ROUGE-2, and 0.20 in both ROUGE-L and ROUGE-LSum. The low scores achieved might be happened because the amount of data used for both training and testing, the limit put on the hyperparameters, and the method of preprocessing the data for the model to train and test with. A better resourceful environment might improve the performance of this model.
Developed by: Daviadi Auzan Fadhlillah
Model type: T5-small
Language(s) (NLP): English
Finetuned from model [optional]: T5-small
Uses
The aim of fine-tuning this model is to generate an abstractive summarization of online news articles, however this model can be used for other purposes where a summarization is needed.
How to Get Started with the Model
Use the code below to get started with the model or to improve the fine-tuning of the model.
