Skip to content

avithekkc/news-headline-generator

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

News Headline Generator (Abstractive)

Header

Overview

This project seeks to create a model that creates one line abstractive summary of a given news article. This model will help to get a quick idea on what the article is all about without having to read all whole news.

Business Problem

News/Media/Blogging Companies always wants a catchy headlines for the articles they post so that they can get maximum number of clicks. A machine learning algorithm can be used to create a headline based on the context of the article.

Data

The data was collected from various sources like Kaggle and Github. Because of the Computational resource limitation the model was trained on 100K News Article having with input length of 100 words and out length of 16 words.

OSMEN Process

  1. Obtaining data
  2. Scrubbing data
  3. Exploring data
  4. Modeling data
  5. Interpreting results

About Modeling Technique

LSTM Network

Train Vs Test Loss

TRAIN LOSS = 2.89 | TEST LOSS = 3.53 loss

Results

Using Seq2Seq LSTM Encoder Decoder model

ROUGE (1) SCORE : 0.2254 ROUGE (L) SCORE : 0.2110

SPARSE CATEGORICAL ACCURACY : 55% ( TRAIN) and 50% ( TEST)

The WebApp version of the project can be found here - Github

Next Steps

  • Adding Attention Layer.

  • Using Bi-directional LSTM.

  • Using pre-trained Embeddings.

Repository Structure

├── datasets                            <- data used for the analysis.
├── images                              <- All images used throughout the project.
├── data-cleaning.ipynb                 <- Cleaning of data to create final dataset.
├── LSTM.ipynb                          <- Data Modeling.
├── presentation.pdf                    <- Non Technical Presentation.
└── readme.md                           <- README for Quick overview on project.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published