What is Sunflower?
Sunflower extracts the text proper out of similarly-formatted HTML files. When you download an offline mirror of a Web site, the HTML files typically consist mostly of noise, such as advertisements, header, footer, a bunch of links. Sunflower is able to extract most of this noise automatically, with only minimal human intervention.
Sunflower (and its predecessor) has been used to extract many press texts for the National Corpus of Polish.
How it works
Just launch Sunflower and point it at the directory containing your collection. Sunflower will pick one file at random, extract all the strings from it and ask you to pick the strings that are contained in the text proper. You don't have to select all the strings -- typically two or three will do.
Once you are satisfied with the extracted document, click Next. Sunflower will then process another file from your collection, using the extraction rules it has just learned, and show you the result for review. Upon confirmation, the entire collection will be backed up (in case something goes wrong), and then modified in place.
Sunflower is written in Clojure and is buildable using Leiningen. Just say
to produce a standalone jar. You will need to have flyingsaucer.jar in your local Maven repository; get the jar from Flying Saucer's site and do something along the lines of:
mvn install:install-file -DgroupId=flyingsaucer -DartifactId=flyingsaucer -Dversion=8.0 -Dfile=flyingsaucer.jar -DgeneratePom=true -Dpackaging=jar
Name and author
The program is so named because it hulls the essence out of HTML files, similarly to removing hull and extracting the kernel out of a sunflower seed.
Written by Daniel Janus, 2010; available under the terms of MIT License. See the file LICENSE for details.