8/3/2007

Cleaned up known typos in userguide.tex, added comments for things
that need to be fixed but weren't easy to fix.

Proposed overall structure for expanded monograph version
of user's guide: 

Working title: Compositional semantics in the wild

Table of contents:

1. Big picture questions.  What are we answering here?
	- Early design decision in HPSG to look at language as
	a system of signs/pairing of form and meaning.
	- No syntactic analysis without semantic analysis/meaning
	representation.
	- Working with broad-coverage implemented grammars in
	this context means we have to apply this commitment to
	developing meaning representation to all the stuff that
	occurs in the data we work with.
	- The Grammar Matrix represents an attempt to do this
	while being as language independent as possible (facilitate
	MT, downstream processing in general)
	- => bottom-up study of similarities/differences between
	languages at the level of compositional semantics.
	- Commitment to compositionality => bottom-up study of
	what aspects of meaning can be represented compositionally.
	- Mention reversability, monotonic accumulation of information,
	relationship to generator, usefulness for MT
2. Brief sketch of MRS to make the next chapter readable
	- Basically our review of Copestake et al 2001, 2005
3. How do our MRSs differ from dependency structures
	- Motivate with description of specific phenomena
	that Dan will propose; things that we handle and
	represent that go beyond what you can see in e.g.,
	f-structure
	- To the extent that f-structures do overlap with
	what we are representing, review ParGram work as in
	a similar spirit.
	- Implications for evaluation across frameworks
4. Framework independent MRS design
	- What the MRSs look like, independent of the type
	hierarhcy etc of the Matrix
	- To be extracted from what is already written
	- Include description of grammaticalized pred names and sort
	hierarchy
	- Features (variable properties) v. predications
5. API: Sem-I, vpm
6. Construction of MRS within the Matrix
	- Other half of existing sections 3-6
	- Types defined in matrix.tdl
	- How the syntax-semantics interface works
	- Add in linking types now in matrix.tdl
	- Add in semantic side of lexical rule types now in matrix.tdl
7. Specific analyses
	- Description of MRS testuite items (or selection from)
	describing what they illustrate and why we represent
	them as we do.
	- Previous chapters will mention phenomena in
	passing or moderate detail, but will point forward
	to this collection.
	- MRS testsuite items will primarily be drawn from
	English, but use examples from other languages where
	phenomena can't easily be illustrated in English (prodrop, 
	numeral classifiers, ?)
	- Also to include: nominalization, serial verb constructions, 
	causative/inchoative, wa-marked topics in Japanese (?),...


General comments:

Focus will be specific representations in MRS/construction thereof as
developed by LinGO/DELPH-IN (and adopted in the Matrix), but not
directly MRS-in-Matrix.

Try to de-emphasize ERG a bit.  Use ERG examples where they are
convenient/motivated, but try to bring in examples from other resource
grammars as well.

We'll need to redo all the MRSs displayed to show the message-free
version.

To read:

Ash Asudeh & Dick Crouch
Gert's emerging volume on semantics and pragmatics in HPSG
CCG/MRS (John Beavers)
RASP-RMRS
Annette's work on dependency grammar & RMRS

Timeline:

Dan to list relevant phenomena to consider in the near term,
More active work starting in January

Publication goal:

Main idea here is a monograph, but need to consider overlap
with other projects planned.  Keep an eye out for piece to spin
off as journal publication, too.
