Version 1.8.1

• Fixes bug in bintokenizer.py which could lead to wrong disambiguation of words within phrases defined in the [ambiguous_phrases] section in Phrases.conf. This could cause strange parse results.
• Many additions and enhancements in Reynir.grammar.

Version 1.8.0

• Added accusative_np, dative_np and genitive_np properties on SimpleTree instances, to convert noun phrases between cases. This inter alia enables correct declension of noun phrases to be shown when checking grammar in text, cf. the ReynirCorrect package.
• Tuning and enhancement of the context-free grammar in Reynir.grammar

Version 1.7.1

• Changes to SimpleTree parsing schema, including VP and TO nonterminals. See documentation for details.
• Greatly expanded vocabulary, with words such as rímix, blörraður, vegu and skjöldu.
• Support for passing options from the Reynir instance on to the tokenizer.
• Many enhancements and corrections in the grammar.
• Better support for derived parser classes that add their own fragments to the base grammar, such as for queries.

Version 1.6.0

• Many grammar improvements
• Added conditional section support in .grammar files, with sections delimited by $if()...$endif()
• Added several grammar error checking rules in Reynir.grammar
• Added support for domains and hashtags to grammar, as noun phrases
• Lower-priority productions can now be added separately to nonterminals via > syntax, which is useful for grammar error rules
• Several additions to vocabulary, including fertugsafmæli, heljarinnar and hæstlaunaður

Version 1.5.3

• Better handling of complex composites ("dómsmála-, ferðamála-, nýsköpunar- og iðnaðarráðherra")
• More robust handling of impersonal verbs with arguments ("mig dreymdi kött")
• Updated word lists (ordalisti-first, -last, -all.dawg.bin) for compound word builder
• Added NP-COMPANY nonterminal for company names to simplified trees
• Updated GitHub repository URLs to mideind instead of vthorsteinsson

Version 1.5.2

Added several compound words to vocabulary. Grammar fixes. No changes to the package API.

Version 1.5.1

• Fixed handling of family names, especially those that can also be given names (Hafstein)
• Added functionality for the benefit of ReynirCorrect
• Better support for multiple parsers using multiple grammars simultaneously
• Better support for addresses, especially those with a number+letter (Laugavegur 21B)
• Vocabulary fixes, including fix for BÍN error in Landmannalaugar

Version 1.5.0

• Added parent, enclosing_tag(), index, tidy_text and span to SimpleTree class.
• Included ord.auka.csv and systematic_additions.csv files, used when building the vocabulary in ord.compressed.
• Reynir.tokenize() is now the function from bintokenizer.py instead of the one from the Tokenizer package.
• BinErrata.conf file now controls both edits and deletions from BÍN when building the ord.compressed vocabulary file.
• The default tokenization pipeline in bintokenizer.py has been enhanced to support the ReynirCorrect package.
• Better support for parsing and detecting errors in the use of impersonal verbs.
• Added S-HEADING and VP-REV nonterminals to simplified parse trees.
• Added _Sentence.terminal_nodes attribute for retrieving a list of terminal nodes in a parse tree for a sentence.
• Lots of additional support for place names and geographical information.

Version 1.4.1

All abbreviation handling moved to the Tokenizer package. Moved large resources, such as the ord.compressed file, to git-lfs (Git Large File Storage). Made the tokenizer used by the Reynir class overridable.

Version 1.4.0

Improved handling of adjective arguments (tengdur flokknum, viðstödd sýninguna, frjáls ferða sinna). Stricter and more accurate parsing of impersonal verbs; Ég dreymi fílinn no longer parses. Added kind, cat, fl and leaves attributes to SimpleTree. Added config/BinErrata.conf with almost 1.600 corrections to the BÍN fl data, which are applied when building the ord.compressed file.