Xrenner

- Disambiguate lexical entries with multiple possible entity types based on dependencies and proportional entity type probabilities
- Added two optional model files:
- atoms.tab - list of entity texts known to atomic (non-nesting), even if we don't know their entity type
- new_modifiers.tab - list of modifiers allowed to violate no_new_modifiers setting in config.ini, if turned on
- depedit V1.4.3 bugfix for node pairs with two relations (e.g. dominance AND distance)
- First release of rudimentary Coptic model in models/cop/
- New config.ini setting ignore_in_acronym for function words to ignore when dynamically recognizing acronyms (e.g. 'of' in Federal Bureau of Investigations = FBI)
- Improvements to unit tests
- Numerous tweaks and bugfixes

- OOV entity resolution uses join model of dependnecy and morphological affix probabilities
- Dynamic hasa expansion
- Batch mode using glob syntax
- isa_subclass_agreement setting to control use of agreement in isa matching
- Allow optional model files
- RTL HTML output for Arabic/Hebrew
- Big hasa dataset for English from UKWAC
- HTML documentation
- More unittests
- Numerous bug fixes

- Support for explicit speaker annotations in input
- Added hasa table to help resolve possessive pronouns
- Added unit tests
- Improved architecture facilitates imports and allows a loaded model to be run on multiple input files
- Substantial performance improvements by using intelligent precompiled matcher objects for new CorefRule class
- New version of DepEdit with better performance and escaping
- Verbose option gives run time
- Various bugfixes