340 likes | 358 Views
Creating & Testing CLARIN Metadata Components A CLARIN-NL project. Folkert de Vriend Folkert.de.Vriend@meertens.knaw.nl Meertens Institute, Amsterdam 18/05/2010 LREC, Malta. Project partners. Daan Broeder Dieter Van Uytvanck Folkert de Vriend Laura van Eerten Griet Depoorter. Outline.
E N D
Creating & Testing CLARIN Metadata ComponentsA CLARIN-NL project Folkert de Vriend Folkert.de.Vriend@meertens.knaw.nl Meertens Institute, Amsterdam 18/05/2010 LREC, Malta
Project partners • Daan Broeder • Dieter Van Uytvanck • Folkert de Vriend • Laura van Eerten • Griet Depoorter
Outline • What is CMDI? • What is the goal of our project? • How to go from a resource to harvestable metadata? • Findings of the project and future challenges
1) What is CMDI? • CLARIN MetaData Infrastructure (CMDI) is the infrastructure used for descriptive metadatain CLARIN (Common Language Resources and Technology Infrastructure) • Descriptive metadata is used to characterize data resources and tools, to facilitate discovery and management in large (virtual) infrastructures and repositories.
Advantages CMDI • Compared to other metadata infrastructures: • Flexibility • Researchers can decide what metadata fits their needs and use ready made metadata components. • Researchers can also create new metadata components if they want. • Complete Infrastructure: software for metadata modeling, editing, harvesting, exploitation • Still compatible with existing frameworks: OLAC, IMDI, TEI
Basic Component Metadata Modeling Lets describe a sound recording Sample frequency Format Size Technical Metadata …
Basic Component Metadata Modeling Lets describe a sound recording Language Name Id … Technical Metadata
Basic Component Metadata Modeling Lets describe a sound recording Actor Name Age Language Sex Language … Technical Metadata
Basic Component Metadata Modeling Lets describe a sound recording Continent Location Country Address Actor … Language Technical Metadata
Basic Component Metadata Modeling Project Name Lets describe a sound recording Contact … Location Actor Language Technical Metadata
Basic Component Metadata Modeling Project Lets describe a sound recording Location Actor Language Technical Metadata Metadata profile
Main principles behind CMDI • Component approach which is flexible and lets you design your own metadata profile • But semantics need to be declared explicitly by making use of concepts that are stored in the ISOcat registry. This way interoperability can still be guaranteed.
2) What is the goal of our project? Testing of CMDI principles by applying them to existing resources at MI and INL
Resources at MI and INL used • Lexical resources(with proper names, monolingual and bilingual lexica, historical and scientific dictionaries) • Linguistic databases (with syntactical, morphological and phonological dialect variation) • Ethnological databases (containing data about folktales, songs, probate inventories and pilgrimages). • Corpora (spoken and written) • Historical documents (bible texts)
3) Workflow from resource to harvestable metadata instance B Construction of XML metadata profiles for each granularity level present in resource Very basic tool kit for creating schema and instances Harvestable metadata instance Resource A Resource analysis C Add metadata to instances
Let’s apply this workflow to one of the resources in the project • Dynamic Syntactic Atlas of the Dutch dialects (DynaSand) • A linguistic database of speech and text to chart the syntactic variation at the clausal level in 267 dialects of Dutch spoken in the Netherlands, Belgium and North-West France.
A) Resource analysis DynaSAND A Resource analysis Data, information, metadata? Granularity levels?
B) Profile construction B Construction of XML metadata profiles for each granularity level present in resource Use existing components
B) Profile construction B Construction of XML metadata profiles for each granularity level present in resource Use existing components Introduce new components
B) Profile construction B Construction of XML metadata profiles for each granularity level present in resource Use existing components Introduce new components Link concepts in new components to existing ISOCat concepts (ensuring semantic interoperability)
B) Profile construction B Construction of XML metadata profiles for each granularity level present in resource Use existing components Introduce new components Link concepts in new components to existing ISOCat concepts (ensuring semantic interoperability) Introduce new ISOCat concepts (ensuring semantic interoperability)
C: Generate schemas and add metadata to instances B Construction of XML metadata profiles for each granularity level present in resource Very basic tool kit for creating schema and instances C Add metadata to instances
Workflow from resource to harvestable metadata instance B Construction of XML metadata profiles for each granularity level present in resource Very basic tool kit for creating schema and instances Harvestable metadata instance Resource A Resource analysis C Add metadata to instances Use existing components Data, information, metadata? Introduce new components Granularity levels? Link concepts in new components to existing ISOCat concepts (ensuring semantic interoperability) Introduce new ISOCat concepts (ensuring semantic interoperability)
4) Most important findings of the project • CMDI appeared flexible enough for the resources selected at MI and INL: • Many existing components could be reused. • Where this was not possible the framework indeed made it possible to make new components. • This was the case for both IMDI and non-IMDI type of resources. • A very general issue when making existing resources available through a metadata infrastructure (not CMDI-specific), is how to deal with “data, information, metadata distinction” and granularity levels. -> Advice: keep an end user perspective (discovery and management). • Document with best practices will be made available on CLARIN.EU website.
Future challenges for CMDI • Existing ISOCat concept definitions can be too specific or too broad (“birth year” versus “birth date” f.i.). What if too many components and concepts are created and the semantics become too diffuse to be useful? • Will we need increasingly more standardization and “cleaning” effort from ISOCat in the future? • Will we need more ways of encouraging reuse of existing components and concepts? • Should we add success indicators?: “this component is already being used by 1 million satisfied customers!” • Should we make more explicit what the benefits of reuse are?: “all of these great tools can be used on your data too when you reuse components X and Y!”.
Some links • CLARIN-NL components: http://www.clarin.eu/cmd/components/clarin-nl/ • ISOcat data category registry: http://www.isocat.org • Tools for creating CMDI: • XML-toolkit: http://www.clarin.eu/toolkit • Component registry and browser and Arbil metadata editor: http://www.clarin.eu/cmdi