Home
Exploration of Automatic Text Summarization Algorithms
Sometimes, even important documents can be too lengthy to read fully from start to finish. Therefore, it is desirable for a computer program to be able to automatically summarize long walls of text into shorter, more digestible snippets.
There are 3 existing methods of shortening documents:
- Extraction of key sentences/phrases, where a program finds the most important phrases in the text and removes the remaining text.
- Document meaning abstraction, where a program understands and paraphrases the gist of the text
- Human-aided summarization, which requires manual effort to read and rewrite a document
Only the first 2 methods are of particular interest in this project, as they are fully automated, and save the most human time and effort. In addition, this project will focus on recently developed deep-learning techniques, such as neural attention models and sequence-to-sequence learning.
The aim of this project is to explore the viability of computer-automated text summarization techniques through automated standardized testing.
This project was done during Melvin and Joe's 6 months internship (semester 3.1) for Ngee Ann Polytechnic.