AI Text Summarization with Hugging Face

Course details

The Hugging Face AI community aspires to advance and democratize artificial intelligence through open source and open science. In this course, Janani Ravi guides you through using this powerful AI tool to summarize text. Learn how extractive text summarization works, as well as its intermediate representations. Explore evaluation metrics, and find out how to navigate and access information on the Hugging Face platform. Go over logging into Hugging Face, using the sumy library for extractive summarization, and using abstractive text summarization with the Hosted Inference API. Learn about sequence-to-sequence models, transformers, and how to use them in Hugging Face. Plus, dive into using a Hugging Face pipeline to perform actual summarization, fine-tuning a transformer model, and exploring several Hugging Face transformers.

Concepts

0. Introduction

01 - AI Text Summarization with Hugging Face

1. Understanding Extractive Text Summarization

02 - Prerequisites
03 - Extractive text summarization
04 - Intermediate representations for extractive summarization
05 - Evaluation metrics for summaries

2. Performing Extractive Text Summarization Using Hugging Face

06 - Exploring Hugging Face
07 - Signing up for Hugging Face
08 - The sumy library for extractive summarization
09 - Extractive text summarization on Hugging Face

3. Understanding Abstractive Text Summarization

10 - Abstractive text summarization
11 - Abstractive summarization using the Hosted Inference API on Hugging Face

4. Understanding Transformers

12 - Sequence-to-sequence models
13 - Attention in sequence-to-sequence models
14 - A brief introduction to Transformers
15 - Transformers in Hugging Face

5. Performing Summarization Using a Hugging Face Pipeline

16 - Using Colab to work with Hugging Face Transformers
17 - Loading the CNN Daily Mail dataset
18 - Cleaning text data
19 - Generating summaries with Hugging Face Transformers
20 - Evaluating summaries using ROUGE scores
21 - Summarizing text and computing aggregate ROUGE scores

6. Fine-Tuning a Transformer Model for Summarization

22 - Understanding tokenizers
23 - Fine-tuning the T5 small model
24 - Pushing the model to the Hugging Face Hub
25 - Summarizing text using the fine-tuned model

7. Exploring Different Hugging Face Transformers for Summarization

26 - Accessing the BBC dataset on Google Drive
27 - Instantiating and cleaning the BBC News summaries dataset
28 - Generating summaries using Pegasus
29 - Generating multiple summaries and computing aggregate ROUGE scores
30 - Generating summaries using BART
31 - Computing ROUGE metrics for a set of summaries

Conclusion

32 - Summary and next steps