Here are the original ideas for discussion topics as suggested by participants through 7/8/05.
Based on the success of the EMELD conferences at getting interesting work done during scheduled small group discussions, I suggest that some of the time (perhaps in larger increments) be used for organized discussion. Depending on the size of the group, we may wish to break up into slightly smaller crowds. Possible discussion topics, unordered, spanning both general and specific (though I think that the more specific we can be, the more useful the discussion is likely to be)...
- Specific steps we as DELPH-IN can take to
- ... improve collaboration
- ... raise the stature of robust, deep processing in the NLP community
- ... become visible as an organization
- Which applications benefit most from deep processing?
- What research avenues are interesting but presently closed for lack of relevant technology?
And now some further topics suggested by participants:
- Questions on use of latest Matrix
- Further developments in RMRS
- Getting generation to work
- Recent developments and ideas in the Grammar Matrix
- Current developments in statistical modeling (primarily parse selection): particularly refining or retraining models for applications in specific domains
- Integration of symbolic and stochastic processing: Prase selection (integretation into both development and runtime systems), possible ways of integrating statistical scope resolution and resolution of modifier attachment (MRS-based), weighting of transfer rules (MRS-based MT), realization-banking/selection in generation.
- Integration of deep and shallow grammars
- DELPH-IN tools and morphology/phonology
- How to approach potential future participants of DELPH-IN: Should we? To what extent? Under which criteria?
- Development of DELPHIN repository (systems, tools, data, lingware)
- Comparative evaluation
- Multilingual applications, including MT
- Prospects for joint projects
- Partnerships, membership and outreach
- Visibility, joint PR, presence at international events
- DELPH-IN web site
- Future development of deep processing software: licence issues, new features, extension of existing tools
- Documentation of grammars and tools: Efficient ways and standard tests
- Strategies for coverage, new domains, lexical acquisition, robustness and evaluation
- Propogation of our ideas: usage of DELPH-IN, teaching, publication efforts
- Joined work and possible project ideas
- Open source licenses: a common DELPH-IN standard?
- Generation from RMRS (what's planned, what's available)
- Matrix updates: What would the community like to see? What makes it feasible to integrate new Matrix versions? What kinds of community input to Matrix development are likely to be most effective?
- Delph-In infrastructures
- Parallel grammars and MT