v6.1.8

Features added:

• Fix default sentence boundaries to exclude long dashes
• Fix sentence evaluator to read input configuration from correct location

This release can use language packs and configuration files from release 6.1.3.

v6.1.7

Features added:

• Add possibility to configure sentence boundary characters as well as indicate whether a language has lowercase/uppercase letters

This release can use language packs and configuration files from release 6.1.3.

v6.1.5

Change log:

• Update dependencies to recent versions
• Update Java from 8 to 11

This release can use language packs and configuration files from release 6.1.3.

v6.1.4

Change log:

• Fix bug when evaluating parser with pos-tagger

This release can use language packs and configuration files from release 6.1.3.

v6.1.3

Change log:

• Fix tokenization in French and French-UD language packs.

v6.1.2

Change log:

• Replace existing distribution with shaded jars
• Expose TalismaneMain option parser to allow subclasses to add new options

This release can use language packs and configuration files from release 6.1.0.

v6.1.0

Change Log:

• Add current file information to "with_location" template when reading a single file
• French Language Packs: Always join compound words separated by single dashes into a single token
• Make reference.conf independent to avoid errors when incorporated in other libraries
• Simplify project structure: move parent pom.xml to top-level directory

v6.0.0

Change log:

• Make all Talismane outputs serializable (Sentence, TokenSequence, PosTagSequence, ParseConfiguration, ParseTree). To enable this:
  • TalismaneSession was pulled out of all constructors, and replaced by a sessionId.
  • This session id is used to load a talismane session dynamically, directly from the configuration file, in the key talismane.core.sessionId.
  • This changes the way Talismane is called, as shown in the examples.

This release can use language packs and configuration files from release 5.2.0

v5.3.0

Change log:

• Add lower case first word filter
• Rationalize pom.xml structure

This release can use language packs and configuration files from release 5.2.0

v5.2.0

Change log:

• Add support for French Universal Dependencies
• Add possibility to merge lines read from CoNLL input
• Add possibility to split lines when generating CoNLL output