NL-Soar naming conventions

Soar Production Files

Soar production files should follow the following convention :- Comprehension files are predictable in following this convention. Generation and Dialogue are less so.

NL-Soar directory structure

NL-Soar directory structure

Within the project/soar/utc/nl directory are a number of subdirectories containing different variants on the NL model. These are named with their version release label. The most recent is nl9702.

Within each release, there are a number of subdirectories containing the various subcapabilites:

nlc
Comprehension code.
nld
Discourse code.
nlg
Generation code.
lexicon
Lexicon entries for both syntax and semantics.
map
Mapping code.
regression
dumps of the syntactic and semantic tree structures built by the version of NL-Soar contained in the parent directory. This is considered the Gold Standard for future versions of NL-Soar which is used during regression testing.
top
Productions for the top space.
trx
Code for the translation driver. Also under the /afs/cs/project/soar/utc/nl/ directory are the following:

ntd
The NTD-Soar agent and its old PSCM version of NL-Soar from circa March 1994. The subdirectories nlc and nlg are the NL portions, as above. The io_code has been specialized for the NTD agent, incorporating NOVA visual attention and relevant motor skills. The Mosaic documentation focuses on the agent, rather than on the language capability.
ta
A collection of files for the TacAir-Soar agent (?)
sounds
A collection of sound files (compressed) spoken by NASA staff, taken from their audio tapes, to be played by the NTD-Soar agent.
traffic
A collection of sound files (compressed) spoken by WDUQ traffic reporters, recorded unofficially from the radio, to be used in studying the feasibility of a traffic "agent".
Written by Julie Van Dyke (vandyke@cs.cmu.edu), August 1997