LisbonRepositoryDiscussion

Designing and creating a DELPH-IN repository (systems, tools, data, lingware)

LisbonRepositoryDiscussion is the transcript of Discussion I on Thu 2005-08-18 (11:00-11:45) in the LisbonTop meeting.

Moderation: StephanOepen

Scribe: UlrichSchaefer

Stephan introduces motivation for the discussion.

• [http://www.delph-in.net DELPH-IN] is a community effort

• central is the common repository

• internal goal: exchange not only ideas, but also tools, resources, experiences

• external goal: increase visibility, marketing; this is closely related to Discussion IV: [wiki:DelphinVisibility Visibility] in the afternoon

  • attract new contributors and funding
  • give the consortium a visible platform with a name

Start with what's on the current webpage on [http://www.delph-in.net/index.php?page=3 Systems & Grammars]

• Core components

  • [http://www.delph-in.net/lkb/ LKB]: development environment

  • [http://www.delph-in.net/pet/ PET]: fast parser

  • [http://www.delph-in.net/itsdb/ [incr tsdb()]]: evaluation and benchmarking tool for both

• Lingware

  • [http://www.delph-in.net/matrix Matrix] : starter kit for new grammars / bottom-up crosslingual

  • [http://www.delph-in.net/erg ERG] : English Resource Grammar

  • [http://www.delph-in.net/jacy Jacy] : Japanese Grammar

  • [http://www.delph-in.net/mgrg Modern Greek]

  • [http://www.ling.hf.ntnu.no/forskning/norsource NorSource] : Norwegian Grammar

  • German grammar to be available as open source in 2006 (already available for DELPH-IN members now)

  • plus more candidates: Korean, French, Spanish, Catalan, ... ?

• Architecture, language technology

  • [http://www.delph-in.net/heartofgold Heart of Gold] : framework for hybrid processing applications

• Treebanks: [http://redwoods.stanford.edu Redwoods], Eiche, Tiger 700, ...

Questions

• how to organize exchange and delivery
  • communication
    • wiki
    • mailing list(s)
    • external vs. internal communication
    • up-to-date info on ongoing developments (visibility; avoid double work)
    • bugtracking system (scapegoat)
    • second generation developers experiences
    • keep in mind 'consumer-only' users like students, people not immediately working in CL/LT research
  • repository
    • what is missing?

      • common morphology component (finite state) for integration with LKB and PET, e.g. Spanish, Portuguese, German, Japanese: availability vs. usage of internal morphology, LKB optimization/lex. rules problem: more readings than necessary

      • better interface to e.g. ambigous external tokenization/preprocessing

      • machine translation (transfer) support

      • documentation is distributed, partly hard to find, e.g. [http://www.project-deepthought.net DeepThought] deliverables

        • see also follow-up Discussion II: [wiki:DocumentationStrategies Strategies for documentation of grammars and tools]

      • quality assurance; see also Discussion V: [wiki:DistributedTesting Developing a distributed means of testing LKB updates] on Friday

      • ways of obtaining DELPH-IN components

        • release vs. full CVS access (development tracking)

      • domain adaptation support

    • which platforms (operating systems, development environments) to support
      • what OSes are currently used?
      • which systems available via DELPH-IN are currently used?
• marketing

  • is it a good idea to separate into wiki and static 'official' web page? (yes)

  • color and design of the current webpage

  • logo

  • online demos (holistic demonstrator / showcase app / killer app)

  • educational material

  • presentations (e.g. DELPH-IN corporate overview, tools, available resources)

  • target audience:

    • attract new contributors (members)
    • funding agencies/companies (e.g. visibility of already funded projects to potential funders)

  • licensing: see Discussion VI: [wiki:DelphinLicenses Developing a standard DELPH-IN open source license] on Friday