Welcome to Reynir

Til að gagnast sem flestum er skjölun Reynis á ensku. - In order to serve the widest possible audience, Reynir’s documentation is in English.

Reynir is a Python 3.x package for parsing Icelandic text into sentence trees. The trees can then be used to extract information from the text, for instance about people, titles, entities, facts, actions and opinions.

Along the way, Reynir tokenizes the text, finds lemmas and assigns part-of-speech (POS) tags to every word.

An example of a parse tree

Reynir is the engine of Greynir.is

Reynir has been used to parse text from Icelandic news websites since 2015, processing 6 million sentences in 300.000 articles. Its optimized C++ parsing core is fast and efficient enough to parse real-world text according to a context-free grammar for the Icelandic language with over 16,000 productions.

To get acquainted with Reynir, we recommend that you start with the Overview, proceed with the Installation instructions, and then look at the Quickstart. For further reference, consult the Reference section.

This documentation also contains important information about copyright and licensing.

Batteries included

To start using Reynir with Python, you (usually) need ony one command:

$ pip install reynir

No database to set up, no further data to download. The entire Database of Modern Icelandic Inflection (Beygingarlýsing íslensks nútímamáls), with over 6 million entries, is embedded within Reynir in compressed form. By looking up word forms in this database and applying context-free grammar rules (productions) and scoring heuristics, Reynir is able to infer what the most likely lemmas are, how they are inflected in the parsed text, and where they fit in the overall sentence structure.

Reynir is thoroughly documented, and its source code is of course available on GitHub.

Enabling your application

Reynir can serve as an enabling component of applications such as:

  • Natural language query systems
  • Bots and conversational systems
  • Information extraction tools
  • Intelligent search tools
  • Grammatical pattern analyzers
  • Text similarity
  • Author identification
  • Sentiment analysis
  • Content summarization
  • Content category labeling
  • Part-of-speech (POS) taggers and lemmatizers
  • Generation of training corpora for machine learning

About Reynir

Reynir is a project and product of Miðeind ehf. of Reykjavík, Iceland. It is a free open source software project (GNU GPLv3), started in mid-2015 by its original author, Vilhjálmur Þorsteinsson. Its aim is to produce an industrial-strength Natural Language Processing toolset for Icelandic, with the hope of supporting the language on the digital front in times of rapid advances in language technology; changes that may leave low-resource languages at a disadvantage unless explicit action is taken to strengthen their position.

Reynir and associated projects received grants from the Icelandic Language Technology Fund (Máltæknisjóður) in 2016 and 2017.