Highlights! - News Text Summarization with Abstractive (T5 Transformer) and Extractive (Text Ranking) Techniques

Problem Definition

With human lives getting increasingly busy day by day, it is difficult to stay in touch with everything that is happening around the world. People want to stay up to date on the latest news while spending minimal time reading it. In such a scenario, reading the summary or highlights of news stories is a more convenient way to keep up with the news. Reading short summaries of all articles saves time for those who do not want to read the complete article, and if someone needs more information after reading the summary, they can read the articles that interest them.

The goal of this project is to automate the text summarizing of news stories. The system generates a brief synopsis of approximately 60-100 words for a news article that is provided. Models were trained and evaluated using ROUGE (Recall Oriented Understudy for Gisting Evaluation) scores