Skip to content

Latest commit

 

History

History
60 lines (41 loc) · 3.73 KB

readme.md

File metadata and controls

60 lines (41 loc) · 3.73 KB

justext

A java implementation of the justext algorithm, which help to extract article main content from a Html page such as a blog post, an article, justex will remove all the ads, banners, menus of the page and give you the main content.

Demo Application

Go to JustextDemo, run

gradle run

Requirement

  • Windows x64
  • Java 1.6+
  • Authorization to write log files to the working directory

User Guide

  • Enter address of any article and click GO to extract the main content

  • You might override the default configuration of the extractor to test. Click Reset to see the default configuration, and tweak from this. You have 3 configuration editor to try different configuration set

  • The extraction result is displayed in several diffrent WebView

    • Final - result from extractor (page after removing boilerplates)
    • Decoration - remove basic useless content (pre-process) then decorate higlight boilerplate in red, main content in green
    • Original - article original
    • Pre-porcess - remove basic useless content (script, form, style...)
    • Auto-detect Language - result from extractor if enable the language-awareness (the extractor is more strict)
    • Auto-detect Language Decoration - result hilighting from extractor after enabling language-awareness (the extractor is more strict)

  • In each WebView (Final, Decoration..), you can View Page Source, and search inside it with

    • Ctrl+F - Open search bar
    • Enter - find next
    • Shift+Enter - find previous

  • At the bottom of WebView is the status of the extraction process

    • Time spent in extraction process (not the downloading process)
    • Address

  • Each component in the interface is an individual module which communicate with each other via Event Bus, I did not have time to carefully program the synchronisation between components. So you might meet some odd behaviour, but it is not bug. For example: if you click GO 2 times to process 2 diffrent articles, you will see the result of the latest Article which finish the processing, it might NOT the last article that you clicked on GO.

  • Justext is language-independant by default, to enable language-awareness you can give the following config to the extractor

    • language = fr; or language = en; so you tell the extractor which is the language of the article.
    • or autoDetectLanguage = true; So it will try to detect the language of the article
    • If the extractor does not possess the corresponding stopwords dictionary, it will automaticly disable the language-awareness during extraction process. The language-awareness extractor is more strict, it rejects paragraphs which are grammatically meaningless.

Notice

  • This application helps you to quickly test the my Justext library with minimum requirement (no need to git, no need to build..). But It is not warranty to be use the latest version of Justext.
  • This application is heavy (~7Mb) because I take occasion to experiment diffrent swing libraries which might be over-necessary for such a small demo application.
  • But the Justext library is small and it is designed to use in Android. Justext only depends on Jsoup and Guava v15
  • I dev and watching WorldCup 2014 at the sametime, you might found many bug and odd design decision / architect everywhere. Please don't take this as good application design.

About Justext

https://code.google.com/p/justext/wiki/Algorithm

This implementation added more rule and features from the original algorithm. And it is not a finishing work, I will inspire the algorithm and make much more change later.

About Readability

Algorithm used in Safari