HCLSIG/SWANSIOC/Meetings/2009-10-16 Conference Call

From W3C Wiki

Conference Details

  • Date of Call: Friday Oct 16, 2009
  • Time of Call: 11:00am Eastern Time
  • Dial-In #: +1.617.761.6200 (Cambridge, MA)
  • Dial-In #: +33.4.89.06.34.99 (Nice, France)
  • Dial-In #: +44.117.370.6152 (Bristol, UK)
  • Participant Access Code: 4257 ("HCLS")
  • IRC Channel: irc.w3.org port 6665 channel #HCLS (see W3C IRC page for details, or see Web IRC)
  • Duration: ~1 hour
  • Convener: Tim Clark
  • Scribe: Jack Park

Agenda

  • Integrated Model of Discourse - Tudor Groza
  • AOB

Documents

Meeting Notes

Analysis of models of discourse representation. Features: (slide 2) 

course/fine-grained rhetorical structure
relation types
polarity
weights
metadata support
domain knowledge--how tightly coupled or open
 purpose,
evaluation and uptake

Models considered: Harmsze, ScholOnto, De Waard, SWAN, SALT

Possible directions for finding unified model (slide 4) 

balance of features
practicality for uptake
general structured (layered)
course-grained structure--rhetorical blocks
fine-grained structures--discourse elements
relations -- 2 layers, argumentative+coherent, rhetorical

Concrete example (slide 6) 

continue work already done by SWAN/SIOC team
SIOC can structure any kind of document
Discourse elements come from SWAN--> fine-grained
Argumentative and Rhetorical separated
  polarity apparent from ScholOnto presence

This is to start a discussion. What do people think?

Tim: some topics/questions First: emphasis on practicality: what is the purpose of different elements in the model. Can you identify purposes of rhetorical/coherence rlns:

Tudor: start from initial purposes of model (e.g. knowledge bases, extending argumentative networks), rhetorical rlns create structure that externalizes to differentiate argumentation of text

Tim: if you implement this model with software, what's use case or who's users of elements?

Tudor: lower level rlns hidden (high ambition) from user and processed in backgrounds (NLP), upper level in direct contact with user: manual or machine assisted; need too see what is reln between relations--overlapping or complementary to each other.

The point was to start the discussion and see what the opinions are.

Tim: our ontology work came from a different perspective; this is a different and active research area; provides hooks to those who want to do research.

Tim: how would this interact with aTags?

Tudor: didn't spend too much time thinking about that. Hope to talk with Matthias about that.

Tim: definitely a good area for further discussion while at ISWC09

Jack: interested in implementation level

Tudor: linguistic processing tools that creates network of elements, then search for argumentation, present in Cohere or similar tool. Connect discourse elements to domain items, e.g. proteins, user can see argumentative discourse behind each domain item.

Tim: two kinds of evidence: other documents and data sets. Can methods and materials be related to evidence? Course-grained rhetorical structure, one block is methods and materials...

Tudor: different rhetorical blocks can have different meanings, e.g. semweb, mathematical domains, experiments where materials and methods.

Tim: how would you build out this model in practical applications? (need to ground in practical app)

Tudor: tough question! Would at least have to start with agreeing with a basic model before going to practical side: agree on fine-grained structure (double layer), etc.

Tim: perhaps start by agreeing with course-grained structure, maybe do we do double layer? (kindof obvious you need to do that).

Tudor: I'm thinking about fine-grained structure.

Tim: semantic annotation of literature: course grained structure is important to us--e.g. structure of methods and materials.

Tudor: depends on domain; different perspectives, who looks at course-grained? motivations? contributions? One way to see it. Can further decompose into, e.g. materials.

Tudor: we can put together existing course-grained and see if we can use it or do we need to come up with something else; need to look at several domains.

Tim: Next meeting November 6, week after ISWC

Retrieved from "https://www.w3.org/wiki/index.php?title=HCLSIG/SWANSIOC/Meetings/2009-10-16_Conference_Call&oldid=2194"