New LKB-FOS binaries

LKB
1

I’ve just uploaded a new binary distribution of LKB-FOS. It can be download it via the usual link on the LkbFos wiki page. The latest source code is in the repository.

Below are the main user-visible changes. I’ll talk a bit about them at the Bergen Summit next week.

John

  • Fixed error “The value :ALL is not of type (OR STRING CONS (MEMBER :WILD NIL)) when binding :DIRECTORY” when starting [incr tsdb()] podium - the temporary 3-line fix in DELPH-IN Discourse posting is no longer needed.
  • Post-generation chart mapping now works on tree FSes not on edge FSes (this can make a difference when information is percolated top-down).
  • Grammars may now be loaded using the TDL-inspired DELPH-IN configuration file format. Loading can be conditionalised as in Sub-Proposal 1 “Another possible syntax …” in the wiki page GrammarConfigurationRfc. A reasonable number of parameters work the same as in ACE, but unfortunately many others differ.
  • Reduced memory usage in parsing and generation, especially for long sentences.
  • Implemented full forest [incr tsdb()] output (still experimental).
2

Hi @johnca, this is welcome news. Unfortunately I won’t be seeing you all in Bergen this year, so I have one question two questions about this (for now). In the “Another possible syntax”, there is the optional parameter for :config environments, e.g.:

:begin :config LKB.
...
:end :config.

Can you confirm whether that parameter for LKB-FOS is LKB or something else?

Is the behavior such that the LKB processes config entries in those blocks and unspecified blocks (without any parameter) while ignoring those with others (e.g., ACE, FOO, etc.)?

3

Sorry not to see you in Bergen.

Regarding that optional parameter: yes to both your questions. LKB is recognised, and in fact a few other relevant values are as well, reflecting sub-systems / features that are available. For the current LKB-FOS binary these include LKB-FOS itself, and also LUI, MRS and YY. The way this is implemented is by checking against the value of the Lisp global variable *features*.

4

Sorry you couldn’t make it. However do you think you might have time to connect remotely to a unified grammar config SIG session sometime on Thursday or Friday? They haven’t been scheduled yet, so if you could in principle then what would your time constraints be? It would be great to get your input!

John

5

Hi,

Thursday after 3pm would work for me.

But did you mean to ask Michael?