HCLSIG/SWANSIOC/2009-06-05 Conference Call

From W3C Wiki
Jump to: navigation, search

Conference Details

  • Date of Call: Friday June 5, 2009
  • Time of Call: 11:00am Eastern Time
  • Dial-In #: +1.617.761.6200 (Cambridge, MA)
  • Dial-In #: + (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: Matthias Samwald



<Susie> Attendees: Tim, Kei, Anita, David, Susie, Matthias, Jack, Alex, Paolo, Julia, Mary, Tudor, EricP

<matthias_samwald> scribeNick: matthias_samwald

<matthias_samwald> Tim: The wiki page for the meeting today is http://esw.w3.org/topic/HCLSIG/SWANSIOC/2009-06-05_Conference_Call

<matthias_samwald> ... it also contains the slides for the presentations today

<matthias_samwald> --- TOPIC: Presentation by Tudor Groza from DERI ---

<matthias_samwald> Tudor: I am finishing my PhD at DERI

<matthias_samwald> ... I want to present SALT (Semantically Annotated Latex -- also used for other formats besides Latex)

<AlexPassant> ericP: wonderful ! http://www.w3.org/2001/sw/hcls/notes/sioc/index.html

<matthias_samwald> ... slide 4 presents the motivation for our work

<matthias_samwald> ... communication between authors

<matthias_samwald> ... with claims etc.

<matthias_samwald> ... we want to capture details on different granularities: conclusion section, claims, arguments...

<matthias_samwald> ... we wanted to externalize this knowledge via ontologies

<matthias_samwald> .. slide 6... 'standing on shoulders of giants'... vannevar bush, ted nelson, RST introduced by mann & thompson

<matthias_samwald> ... slide 7 explains RST

<ericP> AlexPassant, when sharing that, you can shorten it to http://www.w3.org/2001/sw/hcls/notes/sioc/ , which allows us to rename the index file when we publish it

  • ericP LeeF, see? another satisfied customer

<matthias_samwald> ... text spans, schemas, schema applications, structures

<matthias_samwald> ... slide 9 introduces our work

<matthias_samwald> ...we focused on rhetoric and argumentative descriptions in LaTex and Word

<matthias_samwald> ... framework of three layers: linear structure, shallow annotations, rhetoric annotations (slide 10)

<matthias_samwald> ... slide 11 describes our Document Ontology

<matthias_samwald> ... also captures evolution of document if necessary

<matthias_samwald> annotations should be integrated into text documents via pointers in the document (text should not be re-iterated in datatype properties)

<matthias_samwald> ... slide 12 shows the annotation ontology, re-uses existing ontologies

<matthias_samwald> ... slide 13 shows rhetorical ontology, made up of 3 parts

<matthias_samwald> ... namely argumentation, rhetorical blocks & rhetorical relations

<matthias_samwald> ... we distinguish 'nuclei' from 'satellites'

<matthias_samwald> tim: one thing we discovered when working with scientists: it can be difficult in getting them interested in making these fine distinctions

<matthias_samwald> tudor: i agree, we did two evaluations with users

<matthias_samwald> ... without doing anything, we get linear structure of document

<matthias_samwald> ... indeed, users do not see much immediate benefit in giving annotations

<matthias_samwald> ... we need more immediate reward to get the motivated

<matthias_samwald> ... back to presentation... we were inspired by Anita's ABCDE

<matthias_samwald> ... our argumentation model was inspired by IBIS

<matthias_samwald> ... slide 14 shows an example of how markup looks like

<matthias_samwald> ... 6 months ago we started to work on automatic extraction

<matthias_samwald> ... by means of shallow computational linguistics

<matthias_samwald> ... slide 18... we analyzed various discourse markers (e.g. 'although')

<matthias_samwald> Anita: not all relations have such explicit markers

<matthias_samwald> Tudor: yes, but in our case all had such markers

<matthias_samwald> ... current solution is not complete, but the results we get now are correct.

<matthias_samwald> ... slide 20 -- the first system based on this was 'KonneX', which allows users to search / browse assertions

<matthias_samwald> ... slide 21 -- another application is "sClippy". it extracts shallow metadata from PDFs, including discourse knowledge items

<matthias_samwald> ... we focused on high precision, results not complete

<matthias_samwald> ... sClippy accesses DBLP, users can then also browse graph of co-authors etc.

<matthias_samwald> ... slide 22 -- CORAAL. won second prize at Elsevier Grand Challenge.

<matthias_samwald> ... not directly linked with SALT at moment

<matthias_samwald> ... we are looking forward to alignment with SWAN and SchoolOnto

<matthias_samwald> Anita: I sas Siggi in Crete, we are looking into having a simpler version of SALT that is a bit like ABCDE, we are pretty far along on getting this done, we are also looking into integrating it into EasyChair

<matthias_samwald> Susie: How will you integrate with SWAN?

<matthias_samwald> Tudor: I did an analysis of all models that deal with modelling rhetorics and argumentation.

<matthias_samwald> ... have a high-level overview, need to look into it in detail

<matthias_samwald> tim: there are similarities, but our work was motivated by creating immediate incentives to users.

<matthias_samwald> ... some of the things in SALT do not appear in SWAN because of that

<matthias_samwald> ... we need to look into what benefits the additional constructs in SALT have compared to SWAN

<matthias_samwald> Tudor: There is also large overlap with SchoolOnto

<matthias_samwald> Anita: An abstract is more of a 'teaser' than a summary of a paper.

<matthias_samwald> Tim: we should start to compile a bibliography on that.

<matthias_samwald> http://samwald.info/res/2009_06_05%20aTag%20Scientific%20Discourse%20presentation.pdf

<Susie> Matthias: Talking about the aTag project

<Susie> Matthias: Build strongly on specific questions and needs

<Susie> Matthias: Building thin layer of technology

<Susie> Matthias: aTag stands for associative tag

<Susie> Matthias: Content developed by me and Holger S.

<Susie> Matthias: Here's an example

<Susie> Matthias: There's a bookmarklet that can be stored in a browser

<Susie> Matthias: Go to a web site, highlight and extract sentence of interest

<Susie> Matthias: Tags are neighbors of entities in ontologies

<Susie> Matthias: Connected to Wikipedia

<Susie> Matthias: System helps you to type less, aid in finding correct spelling

<Susie> Matthias: Working to create an inviting experience with immediate benefit

<Susie> Matthias: Snippets converted into HTML

<Susie> Matthias: aTag can be generated manually, and automatically

<Susie> Matthias: Working to convert databases into this format

<Susie> Matthias: Soon will have web application with text mining results exposed in aTag format

<Susie> Matthias: Underneath HTML is RDF formulated with RDFa and together with SIOC and other ontologies

<Susie> Matthias: Easy for existing systems to use because of use of HTML

<Susie> Matthias: Can use various forms of reasoning and querying to find unknown relationships

<Susie> Matthias: Can crawl distributed RDF and RDFa data sources

<Susie> Matthias: Can use faceted browsing to navigate over aggregated data

<Susie> Matthias: Shows example using the AD drug Memantine

<Susie> Matthias: Good ranking of search results in UI

<Susie> Matthias: Goal is to keep it simple

<Susie> Matthias: Very promising results

<Susie> Matthias: Re-using popular vocabularies

<Susie> Matthias: Simple development of interfaces

<Susie> Matthias: Balance between human readable and machine readable

<Susie> Matthias: Synergies with other work in HCLS, e.g. BioRDF and LODD

<Susie> Matthias: Nice tie into work with SWAN too

<Susie> Matthias: Lucky to have SWAN SIOC alignment already

<Susie> Matthias: Lot of opportunity with RDFa

<Susie> Matthias: RDFa may help get data providers interested in RDF

<AlexPassant> matthias_samwald: would be happy to discuss MOAT / aTags alignements

<AlexPassant> since MOAT is about usint URIs in addition of tags

<Susie> Kei: How do you incorporate relations, given that not many standard ones exist

<Susie> Matthias: The entities are the relations

<Susie> Matthias: Look at state of the art in biomedical ontologies (e.g. OBO)

<Susie> Matthias: Most predicates are simple

<Susie> Matthias: Can actually get good answers even without the predicates

<Susie> Matthias: Loose some power by not representing predicates, but gain in simplicity

<Susie> Tim: What's the data model?

<Susie> Tim: Could you present on the formal model during the call in 2 weeks

<Susie> Matthias: Yes, I can present in 2 weeks

<Susie> Matthias: But very simple

<matthias_samwald> http://esw.w3.org/topic/HCLSIG/SWANSIOC/Examples

<Susie> Matthias: We use SIOC

<Susie> Matthias: It's summarized on the wiki page

<Susie> Tim: 2 interesting approaches

<Susie> Tim: Thanks Matthias and Tudor

<Susie> Tim: Hope you'll be on the call in 2 weeks

<Susie> Matthias: Will present this work at the BioOntologies SIG at ISMB

<Susie> Tim: Alex, Paolo - where are we with the IG notes

<AlexPassant> http://www.w3.org/2001/sw/hcls/notes/sioc

<Susie> Alex: Still need to finish a few things on the SIOC note

<Susie> Alex: Hope to finish over the weekend

<matthias_samwald> Susie: you can start by circulating among SciDiscourse group

<Susie> Paolo: Need to add examples to the SWAN note

<matthias_samwald> Paolo: At the moment I am working on the wiki page comparing approaches

<Susie> Paolo: Versions should be online by Monday