BoilerPy3 is a native Python port of Christian Kohlschütter's Boilerpipe library, released under the Apache 2.0 Licence.
This package is based on sammyer's BoilerPy, specifically mercuree's Python3-compatible fork. This fork updates the codebase to be more Pythonic (proper attribute access, docstrings, type-hinting, snake case, etc.) and make use Python 3.6 features (f-strings), in addition to switching testing frameworks from Unittest to PyTest.
Note: This package is based on Boilerpipe 1.2 (at or before this commit), as that's when the code was originally ported to Python. I experimented with updating the code to match Boilerpipe 1.3, however because it performed worse in my tests, I ultimately decided to leave it at 1.2-equivalent.
To install the latest version from PyPI, execute:
pip install boilerpy3If you'd like to try out any unreleased features you can install directly from GitHub like so:
pip install git+https://github.com/jmriebold/BoilerPy3The top-level interfaces are the Extractors. Use the get_content() methods to extract the filtered text.
from boilerpy3 import extractors
extractor = extractors.ArticleExtractor()
# From a URL
content = extractor.get_content_from_url('http://example.com/')
# From a file
content = extractor.get_content_from_file('tests/test.html')
# From raw HTML
content = extractor.get_content('<html><body><h1>Example</h1></body></html>')To extract the HTML chunks containing filtered text, use the get_marked_html() methods.
from boilerpy3 import extractors
extractor = extractors.ArticleExtractor()
# From a URL
content = extractor.get_marked_html_from_url('http://example.com/')
# From a file
content = extractor.get_marked_html_from_file('tests/test.html')
# From raw HTML
content = extractor.get_marked_html('<html><body><h1>Example</h1></body></html>')Alternatively, use get_doc() to return a Boilerpipe document from which you can get more detailed information.
from boilerpy3 import extractors
extractor = extractors.ArticleExtractor()
doc = extractor.get_doc_from_url('http://example.com/')
content = doc.content
title = doc.titleAll extractors have a raise_on_failure parameter (defaults to True). When set to False, the Extractor will handle exceptions raised during text extraction and return any text that was successfully extracted. Leaving this at the default setting may be useful if you want to fall back to another algorithm in the event of an error.
Usually worse than ArticleExtractor, but simpler/no heuristics. A quite generic full-text extractor.
A full-text extractor which is tuned towards news articles. In this scenario it achieves higher accuracy than DefaultExtractor. Works very well for most types of Article-like HTML.
A full-text extractor which is tuned towards extracting sentences from news articles.
A full-text extractor which extracts the largest text component of a page. For news articles, it may perform better than the DefaultExtractor but usually worse than ArticleExtractor
A full-text extractor trained on krdwrd Canola. Works well with SimpleEstimator, too.
Dummy extractor which marks everything as content. Should return the input text. Use this to double-check that your problem is within a particular Extractor or somewhere else.
A quite generic full-text extractor solely based upon the number of words per block (the current, the previous and the next block).
While BoilerPy3 provides extractor.*_from_url() methods as a convenience, these are intended for testing only. For more robust functionality, in addition to full control over the request itself, it is strongly recommended to use the Requests package instead, calling extractor.get_content() with the resulting HTML.
import requests
from boilerpy3 import extractors
extractor = extractors.ArticleExtractor()
# Make request to URL
resp = requests.get('http://example.com/')
# Pass HTML to Extractor
content = extractor.get_content(resp.text)