Update process of the Sequoia resources

Several formats are available they are all built from a the single resource which contains all layers sequoia.deep_and_surf.parseme.frsemcor. The figure below describes the different transformations that are applied to produce other formats.

All files are available from the Gitlab project.

Formats depend on:

• presence of PARSEME-FR and FRSEMCOR annotations:
  • with such annotations: suffix parseme.frsemcor and orange background
  • without: suffix conll (or conllu) and green background
• Deep syntactic annotation:
  • with both surface and deep dependencies: basename is sequoia.deep_and_surf
  • with only surface dependencies: basename is sequoia.surf
• Conversion to Universal dependencies
  • Native dependencies: basename is sequoia
  • Universal dependencies conversion: basename is sequoia-ud

The Graph Rewriting Systems (usable with Grew) are available:

• sequoia_proj.grs: in the Gitlab project (folder tools)
• ssq_to_ud/main.grs: in the Gitlab project (folder grs)

Note about universal dependencies

In order to produce the final files for the UD project, two more steps are needed (not detailed here)

• Adding the misc feature SpaceAfter=No when needed (using udapi tool)
• Splitting into three subfiles dev, test and train