Kevin Kuehler
63ec5719a8
* Run `cargo fmt` in the speedreader root * Add a section to the README about fetching the test data, since it is not retrieved automatically.
Brave SpeedReader
At a high level, SpeedReader:
- Distills text-focused document content from a suitable HTML
- Works on HTML documents before rendering them
- Generates HTML output with no external styling or scripting
- Content styled with Brave-designed themes
Structure
SpeedReader has two methods of deciding whether a page is readable:
classifierthat analyses HTML content of a page and applies a pre-trained classifier to decide whether the page is article-style, text-orientedwhitelistthat uses adblock-style rules that explciitly whitelist pages that should be distilled
Any page that is readable should then be passed to an instance of SpeedReader
. Internally it uses the whitelist to determine whether to apply streaming,
rule-based rewriting (using lol_html
to support output streaming), or heuristics approach loosely based on Mozilla's
Readability and Rust implementation
improved from https://github.com/kumabook/readability.
Running the tests
You need to clone the speedreader test data to the
components/speedreader/rust/lib directory.
$ git@github.com:brave-experiments/speedreader-test-data.git data
$ cargo test