August 18, 2011
8 PDT/10 CDT/11 EDT/16 BST/17 CEST
Attending: Litsa Mitsopolou, Luke Woodham, Daniela Giordano, Valerie Smothers
1 Review minutes of last meeting
The minutes were approved.
2 Overview of Learning Resource Metadata Initiative
Valerie reviewed that LRMI is an effort by creative commons to create a metadata specification for data included in the web page to make certain details about learning resources more easily discoverable by search engines. Google, yahoo, and bing are collaborating on schema,org, an effort to use microdata in websites to enhance search and discovery. LRMI is a similar effort funded by the gates foundation.
3 mEducator update
Daniela commented that they are refining parts of the schema. They have turned to the RDF approach. With development of the solutions for sharing content, many things have been implemented. They have included a multi lingual approach in schema. In the last version they included this aspect.
Litsa went on to say that after May there were several changes. They are evolving towards an ontology. Now they can express multiplicity and multilinguality in the ontology. They can translate an instance into another language. All mandatory fields can be translated. In addition, some classes have been restructured. They are also implementing a class to connect with attention metadata. From an instance point of view, they try to use linked data. Instead of writing two character string for en for English, they use a URI to represent the namespace. They have worked on attention metadata – they want to support recommendation of resources, recommend queries, etc. Also include ranking in searches.
Daniela added that once you accept that you want to deal with collections of items, it may be useful to handle classifications directly into the schema. Once you incorporate, you may offer more to the user. Collection may have a recommendation algorithm. Including that will allow us to understand if a collection was created manually or was result of an algorithm classifying. That is the idea of the collection and classification notion. It may be that as collateral action they are delving into process of evaluating the solutions.
Valerie asked about the use case for collections.
Daniela commented that they are operating on metadata. Algorithms have flexibility of letting the user choose what metadata fields are to be entered in the classification process. Out of 20, only 10 may be of interest. Then you obtain groups of resources that have similarities. Once you are interested in a resource, what other resource belongs to the same class? It is a way to broaden the scope of the system through exploratory search. You can discover resources that if you go strictly by keywords you may not discover. It’s an additional way to make retrieval more robust. It is similar to related resources.
4 evip update
Luke commented that additional metadata fields were implemented within evip. That included age and sex of the virtual patient. There was a lot of discussion as to whether that was metadata or virtual patient data. They remained in virtual patient data as well.
Daniela commented that in meducator they have not taken the approach of coding so specifically. In schema, you cannot express age. They use open description. If age comes up, it’s likely it will be retrieved in the search. They are making the system more robust to compensate. One reason is that we are dealing with heterogenous resources. In evip, it was homogenous.
Litsa added that they have many different types of resources. It is a challenge to find commonalities. They tried to be more general. It isn’t that user friendly to have so many optional fields for specific resources. That can be inferred from the description.
Daniela added that they are creating guidelines for metadata creators. If there is not explicit age field, but you contribute case study where age is important, in the description you should mention that. Pascal is working towards that.
Luke agreed most vps probably do have that in the description. For the evip website you can see the metadata. For fields coming from vpd data, they harvest from there. He added David Davies is the person to ask about what metadata fields are useful. There may not be a lot of data on what fields people search for. Metadata fields used did vary from partner to partner. Generally, nothing was deemed mandatory.
Valerie asked if we need to add any elements to healthcare lom?
The group moved to a conversation on repurposing. Litsa commented that in evip they represented repurposing from a system, cultural repurposing, etc. They were talking about content packages, which weren’t necessarily repurposed from anything. The approach is slightly different than meducator. They were able to limit the scope much more.
The groups discussed issues of consent. There is a very simple representation of patient consent relating to patients themselves as well as media included. That had been accepted as fairly simplistic. Andrzej has written about the complications of referencing in metadata. Other fields were relating to educational purposes, that is convered by some fields in healthcare lom but they added extra fields about educational scenario and whether the vps had already been used. Some general and could be considered for healthcare lom. In terms of how they were used, he doesn’t have the data. A lot probably were not used.
Valerie agreed to get back in touch with David.
5 Potential extensions to version 2 healthcare lom
6 Open discussion