A Summarization App for your Pocket Queue
Link to TL;DR | username: test , password: test
Final project for General Assembly's Web Development Immersive. A Node.js & React.js application that allows a user to connect their Pocket account and auto-summarize / compress the number of sentences in the article for faster reading. Uses TF-IDF (term frequency - inverse document frequency) to determine the relative importance of sentences.
Articles URLs are scraped and individual sentences within the article are given a score via TF-IDF. Use the slider at the bottom of the summary to adjust the threshold and compress / expand the article (left for compressed, right for expanded). When expanding, sentences appear by next-most-important score.
- The URL scraper doesn't work well for all URLs
- Summarization works better for harder news articles, and less well with op-eds or narrative pieces that require more sentence-to-sentence context
- Node.js
- Express
- Javascript (ES6)
- React.js (ES6)
- Underscore.js
- Passport (User Auth)
- Pocket API OAuth
- Mongoose (MongoDB)
- Modified Node-TLDR module
- Summarization using TF-IDF (term frequency - inverse document frequency) algorithm
- Summarize by URL
- Connect your Pocket Account & summarize articles from Pocket
- Use range slider to filter sentences by importance
- Account management using Passport
- Authentication validation / flash messages
- Ability to archive Pocket articles
- Implement a better web scraper
- Implement a better TF-IDF algorithm