andiksetyawan / trafilatura

Web scraping library and command-line tool for text discovery and extraction (main content, metadata, comments)

trafilatura: Web scraping tool for text discovery and retrieval

Description

Trafilatura is a Python package and command-line tool which seamlessly downloads, parses, and scrapes web page data: it can extract metadata, main body text and comments while preserving parts of the text formatting and page structure. The output can be converted to different formats.

Distinguishing between a whole page and the page's essential parts can help to alleviate many quality problems related to web text processing, by dealing with the noise caused by recurring elements (headers and footers, ads, links/blogroll, etc.).

The extractor aims to be precise enough in order not to miss texts or to discard valid documents. In addition, it must be robust, but also reasonably fast. With these objectives in mind, Trafilatura is designed to run in production on millions of web documents. It is based on lxml as well as readability and jusText as fallback.

Features

• Seamless parallelized online and offline processing:

  • Download and conversion utilities included
  • URLs, HTML files or parsed HTML trees as input

• Robust and efficient extraction:

  • Main text and/or comments
  • Structural elements preserved: paragraphs, titles, lists, quotes, code, line breaks, in-line text formatting
  • Extraction of metadata (title, author, date, site name, categories and tags)

• Several output formats supported:

  • Plain text (minimal formatting)
  • CSV (with metadata, tab-separated values)
  • JSON (with metadata)
  • XML (for metadata and structure) and TEI-XML

• Link discovery and URL lists:

  • Support for sitemaps and ATOM/RSS feeds
  • Efficient and polite processing of URL queues
  • Blacklisting

• Optional language detection on extracted content

Evaluation and alternatives

For more detailed results see the evaluation page and evaluation script. To reproduce the tests just clone the repository, install all necessary packages and run the evaluation script with the data provided in the tests directory.

500 documents, 1487 text and 1496 boilerplate segments (2020-11-06)
Python Package	Precision	Recall	Accuracy	F-Score	Diff.
justext 2.2.0 (tweaked)	0.870	0.584	0.749	0.699	6.1x
newspaper3k 0.2.8	0.921	0.574	0.763	0.708	12.9x
goose3 3.1.6	0.950	0.629	0.799	0.757	19.0x
boilerpy3 1.0.2 (article mode)	0.851	0.696	0.788	0.766	4.8x
baseline (text markup)	0.746	0.804	0.766	0.774	1x
dragnet 2.0.4	0.906	0.689	0.810	0.783	3.1x
readability-lxml 0.8.1	0.917	0.716	0.826	0.804	5.9x
news-please 1.5.13	0.923	0.711	0.827	0.804	184x
trafilatura 0.6.0	0.924	0.849	0.890	0.885	3.9x
trafilatura 0.6.0 (+ fallbacks)	0.933	0.877	0.907	0.904	8.4x

External evaluations:

• Most efficient open-source library in ScrapingHub's article extraction benchmark.
• Best overall tool according to Gaël Lejeune & Adrien Barbaresi, Bien choisir son outil d'extraction de contenu à partir du Web (2020, PDF, French).

Usage and documentation

For further information please refer to the documentation:

• Installation
• Usage: On the command-line, With Python, With R
• Core Python functions
• Tutorials
• Evaluation

License

trafilatura is distributed under the GNU General Public License v3.0. If you wish to redistribute this library but feel bounded by the license conditions please try interacting at arms length, multi-licensing with compatible licenses, or contacting me.

See also GPL and free software licensing: What's in it for business?

Roadmap

• [-] Duplicate detection at sentence, paragraph and document level using a least recently used (LRU) cache
• [-] URL lists and document management
• [-] Configuration and extraction parameters
• [-] Graphical user interface
• [ ] Interaction with web archives (notably WARC format)
• [ ] Integration of natural language processing tools

Contributing

Contributions are welcome!

Feel free to file issues on the dedicated page. Thanks to the contributors who submitted features and bugfixes!

Author

This effort is part of methods to derive information from web documents in order to build text databases for research (chiefly linguistic analysis and natural language processing). Extracting and pre-processing web texts to the exacting standards of scientific research presents a substantial challenge for those who conduct such research. Web corpus construction involves numerous design decisions, and this software package can help facilitate text data collection and enhance corpus quality.

• Barbaresi, A. "Generic Web Content Extraction with Open-Source Software", Proceedings of KONVENS 2019, Kaleidoscope Abstracts, 2019.
• Barbaresi, A. "Efficient construction of metadata-enhanced web corpora", Proceedings of the 10th Web as Corpus Workshop (WAC-X), 2016.

You can contact me via my contact page or GitHub.

Going further

Online documentation: trafilatura.readthedocs.io.

Tutorials: overview.

Trafilatura: Italian word for wire drawing.

Corresponding posts on Bits of Language (blog).