Local Build Instructions

Recipe for Stephan to obtain a collection of JAR files in the right directory structure to populate the LAP Tree.

Conventions Assumed

• no discontinuous regions
• no annotations on edges

Observations and Open Questions

• ? uniqueness of identifiers
• currently unable to run the same tool twice
• record provenance:
• why is ‘annotations’ field a dictionary? by convention never more than one annotation; the key seems redundant
• similarly, ‘links’ is list of lists for LAF compliance
• ‘index’ encodes sequence information but not absolute positions

Action Items

• make LXF output proper JSON: hide _id and produce double-quoted JSON strings
• hide ‘head’ on dependency nodes
• hide ‘annotations’ on edges
• include --threads and --identifier as command line options
• pass tool identifier as parameter into adaptor

Hypothetical (For Now)

• what should be the graph structure when a dependency parser annotates two separate layers of token-level analysis, e.g. PoS tags and stems (as features)