louismullie/treat

Treat - Text Retrieval, Extraction and Annotation Toolkit

1.0.0 - First stable API.

1.0.1

 * Major fix: bug when using the Stanford bindings due to an API change.
 * Final touches to make decision tree classification work neatly.
 * Add method in Countable to calculate position from end of parent entity.
 * Assume HTML format in Buildable when given a URL with no file extension.
 
1.0.2

  * Major fix: fixed the unserialization of "special" values (boolean, symbol, numeric).
  * Minor fix: bug in Iterable#entities_with_feature fixed.
  * Minor fix: bug in Countable#position_from_end fixed.
  * Began AI4R integration and added a simple wrapper for the Multi-Layer Perceptron.
  
1.0.3
  
  * Feature addition: allow to configure the paths to external Stanford files.
  * Major fix: Made entities_with_feature use get() instead of send().
  * Small API change: Renamed Clitic to Enclitic.
  
1.0.4
  
  * Major fix: fixed Stanford tagger & name tagger paths.
  
1.0.5

  * Added French support and multiple tag set support for Stanford tools.
  * Fixed decimal point escaping across whole library.
  * Major removal: merged the :perl tokenizer into the :ptb tokenizer (similar regexps).
  * Added Reuters models to installer.
  * Renamed Lingustics class to Universalisation.

1.0.6

  * Made general topic extraction available on documents, zones and sections.

1.1.0

  * Complete refactoring of the core of the library.
  * Separated all configuration stuff from dynamic stuff.

1.2.0

  * Added LIBSVM and LIBLINEAR classifier support.
  * Added support for serialization of documents and data sets to MongoDB.
  * Added specs for most of the core classes.
  * Several bug fixes.

2.0.0rc1

  * MAJOR CHANGE: the old DSL is no longer supported. A new DSL style using
    lowercase keywords is now used and must be required explicitly.