Best Practices - Multilingual Linked Open Data -- 22 May 2014

<scribe> scribe: daveL

<MTCarrasco> +MTCarrasco

<jgracia> agenda: https://www.w3.org/community/bpmlod/wiki/Teleconference,_22/05/2014#Agenda

roberto asks: how to contribute to wiki

jose: log in with w3c account then your are free to edit

jorge: split call into two parts
... first review list of patterns for describing multilingual information
... second guidelines on the generation of ML resources in linked data
... including migration of existing ML resources to linked data

<jgracia> https://www.w3.org/community/bpmlod/wiki/Best_practises_-_previous_notes#Patterns_for_Textual_information

patterns for textual information

jose: review patterns documented on wiki

jorge: label everything pattern, seem generally useful

philip asks: are these rdf labels, or labels in any vocabulary

jose: agrees perhaps be specific to rdf labels

philip: suggest best practice is RDF labels, and then add other labes, e.g. skos, as preferred, but minimum should be RDF labels

philipp: issue is that tool can easily explode labels, so readers don't know which to look for
... to specify rdf label as minimum

jose: agrees

chaals: benefit of using rdf label is any tool will pick it up
... so use rdfs labal or something specifically subclassed from rdfs
... but not your own label name.

<labra> subproperty instead of subclass

chaals: people using their own should be subproperty of rdfs label

<jmccrae> http://www.w3.org/TR/2014/NOTE-ld-bp-20140109/#multilingual-vocabularies

jmccrea: points to existing w3c guidleines on use of label in ML vocabs

jorge: suggest rdfs labels should be specified here as the best practice

<labra> possible name: use rdfs:label for everything

<jgracia> ACTION: jose to update p2.1 for including rdfs labelling [recorded in http://www.w3.org/2014/05/22-bpmlod-minutes.html#action01]

<trackbot> Error finding 'jose'. You can review and register nicknames at <http://www.w3.org/community/bpmlod/track/users>.

<scribe> ACTION: labra to update p2.1 for including rdfs labelling [recorded in http://www.w3.org/2014/05/22-bpmlod-minutes.html#action02]

<trackbot> Error finding 'labra'. You can review and register nicknames at <http://www.w3.org/community/bpmlod/track/users>.

<scribe> ACTION: jlabraga to update p2.1 for including rdfs labelling [recorded in http://www.w3.org/2014/05/22-bpmlod-minutes.html#action03]

<trackbot> Created ACTION-1 - Update p2.1 for including rdfs labelling [on Jose Emilio Labra Gayo - due 2014-05-29].

pattern p2.2, multilingual labels

jose: not enough to add labels, but also need language tag

<labra> SELECT * WHERE { ?x ex:position "Professor" . }

<jgracia> jose: a drawback for using lang tags is that sparql need to know the language

john: this sounds more like an anti patterns?
... applications that are language dependent should be specifying language anyway

<labra> this can be solved as: SELECT * WHERE { ?x :position ?p . FILTER ( str(?p)="Professor" ) }

john: should express it as an anti pattern not to use language tags

<jgracia> p2.3

<labra> next pattern could be that anti-apttern

jose: agrees

jorge: but you are discussing P2.3, as an anti pattern

chaals: if you don't specify language of literal do you not get the string

john: the language tag makes the literal different so it won't be matched directly as a literal
... but you can search without language tag using FILTER

chaals: the problems with calling this an 'anti pattern' which is it matches behaviour of nieve user
... so need some more guidance on the type of sparql query to use to deal with how people do the 'wrong' thing
... so need to be more explicit on why we need this anti pattern
... happy to remove if we can show this isn't common bad practice

<labra> a new pattern on how to do sparql queries on multilingual linked data?

jorge: suggest perhaps we don't need anti pattern and just promote good practice

jose: should be clearer of what is not best practice

<jgracia> I suggest to add the "how to do sparql" in a "guidelines" document (not the "patterns" one)

jose: and include patterns on how consumer deal with poor practice in sparql queries

jorge: suggests such consumption guidelines on use of sparql could be in a spearate document

<jgracia> https://www.w3.org/community/bpmlod/wiki/Teleconference,_22/05/2014#Agenda

jorge: finish topic on patterns here and continue in later call

series of guidelines generation of ML Language Resources

jorge: elaborate guidelines for generating ML language resources
... such as corpora, terminologies etc
... this can build on work by john and others
... but capturing best practice
... now discuss a proposed template for such reports
... see ToC on agenda

tiziano: have just sent guidelines on converting babelnet to linguistic linked data

https://docs.google.com/document/d/184C_AjY7_PYBSc8SnAFghGLyTo1v312N34dsP9QZokI/edit#

<jgracia> john: document what to do after publishing (versioning, etc)

roberto: versioning are important
... better to start with specific reports, then produce more generalised guidelines

<jgracia> bottom-up approach

roberto: we can restructure the draft above to your ToC

jorge: propose create reports and collect who is interested in contributing to which ones

dave: asks about alignment with ld4lt work on a LR meta-data onto based on META-SHARE

jorge: ld4lt should develop this model and this group develops the guidelines to use it

dave; asks about using w3c report templates

jorge: is exploring this

jose: w3c report template, just involves editing a HTML template

<labra> http://bpmlod.github.io/report/

<flati> w3c html version like this: http://bpmlod.github.io/report/ ? Seems like duplicated information as in https://www.w3.org/community/bpmlod/wiki/Best_practises_-_previous_notes

roberto: happy to move current doc to this w3c format

<jgracia> Resource types:

<jgracia> - wordnet

<jgracia> - babelnet

dave: volunteers to look at translation memory/parallel text

<jgracia> - bilingual dictionaries

<jgracia> - corpora

<chaals> [adjourned]

<chaals> s/scribe: jmccrae//

- DRAFT -

Best Practices - Multilingual Linked Open Data

22 May 2014

Attendees

Contents

patterns for textual information

pattern p2.2, multilingual labels

series of guidelines generation of ML Language Resources

Summary of Action Items

Scribe.perl diagnostic output