This page is just a place to keep the notes we took in creating the discussion topics for the Fefor meeting.

Discussion Topics

A compilation of the topics proposed by participants. Overlapping topics have been merged. Some of the topics looked more requests for talks by others, and have been categorized as such. This classification gives an initial proposal about which session to put each discussion into. No attempt has been made to align this with FeforProcessors yet. [ see AAC notes ]

For Plenum

  • /MRSs in translation and the prospects for MRS-based interlinguas vs MRS-based transfer (Lars Ahrenberg; volunteer) [ AAC - better in FeforRmrs? ]
  • interfaces to ontologies (Anette Frank) [ AAC - FeforRmrs ]
  • finer-grained lexical semantics (Anette Frank) [ AAC - FeforRmrs ]
  • /Whether to propose an ACL SIG on deep linguistic processing (Emily Bender)
  • FSPP (simple preprocessor) module, usage in specific grammars; preprocessors; preprocessor implementation and integration (Ben Waldron; Francis Bond; Ulrich Schaefer) [ AAC - FeforProcessors ]
  • large lexica and LexDB (Ben Waldron; Francis Bond) [ AAC - FeforProcessors ]
  • /Lexical acquisition, esp. for immature grammars (Emily Bender)
  • automated testing (Francis Bond; Ulrich Schaefer) [ AAC - initial session in FeforProcessors ]
  • grammar interfaces: clear interface for robustness-related grammar elements (paths/types/pos-tags/expected tokenization/punctuation etc.) e.g. specified in the inline documentation in such a way that it could be processed/extracted automatically from grammar sources (Ulrich Schaefer) [ AAC - initial session in FeforProcessors ]
  • /how strong is the desire for discontinuous parsing as an option? an implementation could actually be straightforward, but what is needed in terms of the formalism and processing efficiency? (Stephan Oepen)
  • /how do grammarians and developers feel about a limited library of relational constraints? would we know how to provide, say, basic list and maybe set functionality? (Stephan Oepen) [lists yay sets ick –EB]
  • we should review the ToDo list(s) from previous meetings and make sure we maintain a shared inventory of known issues and candidate extensions (Stephan Oepen) [ AAC - see OpenissuesTop proposal ]

Candidates for discussion topics in Grammarians session

(R)MRS details

[ AAC - see FeforRmrs ]

  • MRS design issues
    • Pondering the constraint that all noun-relations must have an associated quantifier, in light of languages that don’t ever (or hardly ever) express quantifiers overtly within NPs (Emily Bender)
    • Inserting pronoun-rels v. leaving unbound indices in pro-drop (Emily Bender)
    • encoding of information structure and general discourse-level bits of information; context modeling in (R)MRS: anaphora, …; reprsentations of discourse status that improve on current def_q_rel mess (Stephan Oepen; Anette Frank; Emily Bender)
  • MRS well-formedness conditions (Ben Waldron)
  • messages: what are they good for? (Francis Bond)
  • the status of the HOOK.INDEX in MRS/RMRS (Francis Bond)

Other

  • Best practices for documenting grammars (Emily Bender)
  • questions/requests that have emerged out of Sara’s work on the BiTSE grammar, and may be linked to priorities in the development of the Matrix, such as empty elements, trigger rules for empty elements, and support for lexemes with more than one semantic relation (Lars Ahrenberg)
  • Evaluation: for instance, in Lisbon one proposal in relation to semantics was to have a test suite of sentences created, which would then be translated into all the languages currently in use in DELPH-IN, in order to be able to have a better overview of the semantic representations and see whether these could be further used in parallel in applications; design, availability and uses for cross-linguistic evaluation resources, particularly semantics-based resources (Valia Kordoni; Laurie Poulson) [ AAC - MRS test suite in FeforRmrs ]

Candidates for discussion topics in Processors session

see FeforProcessors

  • LKB/PET compatibility issues (Ben Waldron)
  • Shared configuration between LKB/PET: reducing and documenting the globals (Francis Bond)
  • standoff annotation interface standard (Ben Waldron)
  • coordination/discussion on SMAF (Ulrich Schaefer)
  • PET API: proposal and discussion (Ulrich Schaefer)
  • for the german grammar, chart dependencies are used heavily but the current LKB and PET behavior differs. maybe we should try to identify an appropriate generalization and implement it (Stephan Oepen) Please see discussion on developers list - AAC.
  • with the german grammar, ambiguity packing is far less effective than for the ERG (or JaCY, for all i know); berthold would like to see a richer restrictor mechanism. related to this, in generation he cannot combine packing and index filtering correctly, because some PVP fronting cases (seemingly) obscure indices too early (Stephan Oepen)
  • both berthold and dan have been interested in modifier attachment underspecification for a while; i expect a good solution may need support in the processors and new types of MRS post-processing; can we set things up for generation from underspecified inputs? (Stephan Oepen) [ AAC - FeforRmrs ]

Correspond to Talks

  • snapshot of where the different grammars are at, in terms of lexical items, lexical types, etc., but also development of treebank data, generation, production of MRSs, compatibility with the Matrix, etc (Tim Baldwin)
  • multilingual grammar engineering issues (Ben Waldron) [? Not sure what this means –EB]
  • Matrix plans (when will there be a new release? what will it contain?) (Francis Bond; Valia Kordoni)
  • how morphology is handled in the different languages; morphology module; Morphology: recent developments in LKB; treatment of morphology in the other grammars, especially the “newer” ones (Tim Baldwin; Ben Waldron; Valia Kordoni)

Would welcome tutorial on:

[ see FeforProcessors - AAC ]

  • Heart of Gold (Tim Baldwin, Lars Ahrenberg, Francis Bond, Stephan Oepen)
  • Lexical Database (Tim Baldwin, Lars Ahrenberg)
  • [incr tsdb()] (Lars Ahrenberg)
  • PET (Lars Ahrenberg)
  • DELPH-IN web demo (Francis Bond, Emily Bender)
  • SEM-I (Emily Bender)

Willing to provide tutorial on:

[ see FeforProcessors - AAC ]

  • how to define SDL cascades in HoG or general HoG intro or ‘installation party’ (Ulrich Schaefer)

Last update: 2006-06-01 by EmilyBender [edit]