Jul 15, 2009
Originally posted on Half an Hour, July 15, 2009.
From the IMS Global Meeting in Montreal, summit on Standards for Pedagogically Relevant Learning Environments.
Where Are we?
Gilbert Paquette
The main reason we do repositories is that we think we can do better quality of learning resources.
The key connecting standards we are working on include Common Cartridge, Learning for All, Learning Design, etc. Common Cartridge aims to emcompass SCORM. There is a common content package, a metadata standard - we will talk about Dubmin Core, LOM, and ISO-MLR (a new standard).
For Common Cartridge sequencing IMS is looking at Sinple Sequencing, but also at Learning Design, and possibly how to join the two. In the future, IMS wants to implement progressively in Common Cartridge other standards for different goals, eg. Content Authorization Standard.
When we look at all of these standards, lots of them involve interoperability, esp. on different platforms. These do not change dramatically the quality of learning. Standards we will talk about later - Digital Repositories, Learning Design, will have more impact.
The first subject, then, is GLOBE - a network of learning repositories.
Why are we interested in learning object repositories? First, because they support resources maintained by institutions, and professors maintain their qualify through peer review and actual use. Additionally, the metadata provides valuable information about authors, subjects, etc. This enables focused queries. And the vast majority of material will be reused.
The first panel:
- Gilles Gauthier - convener at ISO SC36/WG4
- David Massart - LODE co-leader and ASPECT project manager
- Gilbert Paquette - chair (Tele-university)
Questions: are the days really over for 'one metadata for all'? What about the multiplication of profiles? Should we promote a new interchange method? Is there a need for a new standard encompassing other standards? What about the new standards, such as ISO-MLR? New projects, such as competencies?
Gilles Gauthier - Metadata for Learning Resources
The MLR is the standard development that is going on at ISO and is almost ready. It's not a big standard. Many of the people in this room work on ISO and are editors of this standard.
Gilbert asked questions. First, are the days of 'one standard' really over? My answer is 'no'. One standard is nice, if we take out the RDDF triples. Is the multiplication of profiles a problem? No. Is there a need for a new standard encompassing all approaches? Yes?
ISO/IEC JTC 1/SC 36/WG 4
ISO - International Organization for Standardization - is composed of technical committees; there are hundreds of them. The JTC is a joint technical committee with IEC - Intrnational Electtrotechnical Commission. JTC addresses standardization in the field of information technology. There afre different types of members - participating, observing and liasion.
Participating members are each representatives of countries, only one vot per country. Subgoups, called Working Groups, can also include experts nominated by participating members. There is a very clearly defined process - proposal prepatory, committee (where the voting starts), approval, and publication. The idea is to achieve consensus on a technical document.
SC36 is the committee on learning technology. SC37 is in biometrics. SC2 is the committee that produced US-ASCII and Unicode, etc. Under SC36 there are WGs (WorgGroups) including different learning technology standards. The members of SCC include national federations. Liasions include organizations like AICC, ADL, etc.
Working Groups have projects; WG4 has 9 projects, 6 related to MLR and 3 related to IMS content packaging. SCORM 2004 3rd edition became a technical report from this group.
Projects:
MLR 1 - framework. This was the hardest, defining the process for the rest.
MLR 2 - core elements. Essentially, this is Dublin Core.
MLR 3 - core application profile. This is an application profile for the core elements.
MLR 4 - technical elements
MLR 5 - educational elements
MLR 6 - availablity, distribution and intellectual property elements
CP 1 - information model
CP 2 - XML binding
CP 3 - best practice and implementation guide
Various resolutions enabled this work. One major resolution was to maintain compatibility with IEEE-LOM.
The scope is restructed to specifying, in a rule-based manner, metadata elements and their attributes for the descriotion of learning resources. This is a multipart standard, with different components. Multiple languages and multicultural requirements are supported - eg., you only use neutral identifiers., non-linguistic identifiers. All languages are equal. Also, it is intended to support multiple levels of granularity, and to support user extensions.
From a centralized resource to a distributed resource. Centralized, only one body can provide metadata. But metadata on the web can be provided by anybody. You search for the metadata, collect them, filter them, transform them, store them, explore them. So you can get from those RDF graphs MLR records. The MLR view is the centralized point of view.
The idea is to be able to specify data elements - to define what kind of data you want to have. The central element is the respurce. In the MLR all data is simple - all triples - if you want to have something complex (like a vcard) it means you want to have a resource.
Data element specifications - each has a unique identifier, each specifies exactly the possible values, each has human language equivalencies.
Data elements spec ID:
Subject: resource being described - resource ID
Content value: Fodors
Language value: eng
Resource identifier: ID
name: name
Definition: the definition
SubclassOf: other resources
Note: note
If you have a set of MLR elements, you can create an RDF binding. And if you use an XML binding, you get back something that will look exactly like an IMS-LOM record. The MLR is quite simple: it tells you how to define data elements, and how to define application profiles.
David Massart - Learning Object Exchange
I work for European Schoolnet, a network of 31 ministried of education in Europe. It is dedicated to improving learning through technologies. One of its projects is a learning resource exchange - we want all resources in Europe to be accessible.
The LRE is a service for ministries of education driven by them. With 31 countries, many languages, etc., we want 'content that travels well'. Think of music - whatever the language, it travels well. Compare with a lesosn plan in English, for example.
For the purpose of this discussion, a learning object is any resource thaqt can be used in learning. For any resource,
- you need metadata to describe it, also to assess it.
- You need the information to determine whether you want to get access to or not.
- Then once you decide you want access, you need metadata to help you use the resource.
The LRE garters metadat from ministries of eductaion, publishers, museums, teachers, etc. and makes the aggregated metadata available to LMSs, portals, etc. Various standards and protocols are involved: SOI, SPI, OAI-PMH, etc. We try to get metadata by all means, and to provide it by all means.
When you try to do that, you have to really take care of your metadata. There is a lot of negotiation, we try to agree with a provider as to what metadata must be. We control for quality and sometimes reject it. We need to correct and compile metadata, to complete the metadata records. We use an internal format:
- identifiers
- language blocks, eg. title or description in English
- indexes
We have the problem of identity. When we get a record, we need to determine whether this is a new resource, or one we already have, to update the record. Idenification of resources is key. We need to be able to uniquely identify resource - difficult because of political saspects. If you control the identity of the resource, you control the access to the resource.
IMS Learning Object Discovery and Exchange - deals with how to find repositories. There is a regitory of repositories and protocols. Learning objects can have different versions (English, French), different formats (SCORM, CC), copies, etc. This is important if you need to use an object, you need it in a format you support, in a language you read. All of these are contained in Information for Learning Object Exchange (ILOX).
Discussion
Gilbert: We have a diversity problem - it's difficult to federate, harvest, etc. And it contradicts the idea of freedom (to have one standrd), yet we want to interoperate with others. Let's suppose people are using MLR, or ILOX, how will it solve this problem.
David: I don't see diversity as a problem. We have different communities with different needs. And we are simply trying to meet their needs. You define something that is historic in your community, you start to use it. Then you interact with other communities - you negotiate. Anything that helps express the semantioc diference and express these mappings (as good).
Gilles: the difference is, you have a learning resource, it has different langauges it could be, different formasts it could be - all of those are different resources, with relations between them. Variants. Suppose you have a community that is blind. If you try to centralize them, you have more and more and more. What is a contributor: could be a person, an organization, or a service. More and more. Suppose you have a person. You have first name and last name. But thst doesn't make sense in some places. Let the community define them.
David: what you have then is a frameowrk, how to create elements and how to name them. So you have a nice framwork for identifying things. But at the end, whatever profile you use, you will start to exhange these descriptions.
Gilles: yes and no. The ISO-LOM survey from 2004 - most of gthe LOM elements people do not provide information for. If you go to MERLOT and look at the LOM record. They're mostly empty. What MLR will say, a property must have a domain, and must have a range.
Mike: I've been in meetings like this for 15 years. The purpose of metadata is discovery, wouldn;t you agree? Yes, I can find more information on Google, which uses none of that.
Gilles: As a human bneing.
Mike: as a human being. Behind Google is an incredible amount of machine leaning. Metadata has this problem. We have all thiose strucdtures, but we can't match it with the need. I still can't find what I need! How do we do it more like the way Google does, full-text indexing, etc. I understand your comments about the importance of standards, but I think we're a bit myopic.
David: it's not that you use one metadata standard rather than another. Google doesn't do such a great job when it comes to finding resources. If I want to do a specific query, I'm not sure how to express this query in Google, and I'm not sure how I will gind the resource. At the moment Google doesn't do a very good job, and we will find a better resource.
When we look at what people are soing, what we are looking at, is people who are trying to derive metadata in an automatic way. We see a text in English. We track usgae, and we see some resources travel well, and others don't. We look at the resources that do travel well, and we try to understand why. We say to providers, we are interested in these reosurces. It's a web 2.0 approach - we look at tags, bookmarks, keywords, descriptors, etc. We see uses that we didn't forsee.
The key of the ILOX approach is that it's a container. We try to encapsulate metadat created by providers, but also taxonomies, ratings, etc., generated by users. We don't make any assumption on what is a useful metadata or not.
Gilles: I also agree, Google, for human beings, you will find something, but not much use for machines. I would liek to see, just another way. You have a resource, and you have an identifier. You can have as many identifiers as you want, you can go to the international identifiers union and get a unique ID for free. I would like to ask Google, this record I want to have, and then there is a graph of related resources, this resource is the blind, etc. Google could do a better job presenting links.
Comment: Good, MLR aligns with the semantic web, which is where it has to go. Now, as an abstract model, it sounds a lot like the Dublic Core abstract with the semantic web. All of those seems to align one-for-one seem to align with those movements. How does MLR differ from them?
Gilles: Global ID is the same. What we add is a ay to specify an applucation profile to get LOM-like records. Also, the way this is multilingual, using a neutral identifer. All those data elements will be on an international standard.
Comment: when you talk about the URIs, are you just saying they are opaque, that they don't carry semantic data?
Gilles: they could. It's just the way you create your URIs. They have structure. You have rule sets. This is very precise, to speify the domain, the range. This is a namepace for ISO. ISO has a whole structure. We will use that.
David: The problem being if people assign different unique identifiers to the smae resource.
Gilles: You won't have that. You will have one.
Gilbert: you've seen two major initiatives to try to reconcile free and centralized descriptions of resources. These two initiatives can probably link together. ILOX can use MLR. And what W3C is puting out.
Diny: we've gone throught he same process with GEM, and 15 million resources with metadata that needs to be massaged. We have transition that store to an RDF store. So we live in both worlds now - the XML and the RDF world. The blending is not only attainable, but we're doing it.
From the IMS Global Meeting in Montreal, summit on Standards for Pedagogically Relevant Learning Environments.
Where Are we?
Gilbert Paquette
The main reason we do repositories is that we think we can do better quality of learning resources.
The key connecting standards we are working on include Common Cartridge, Learning for All, Learning Design, etc. Common Cartridge aims to emcompass SCORM. There is a common content package, a metadata standard - we will talk about Dubmin Core, LOM, and ISO-MLR (a new standard).
For Common Cartridge sequencing IMS is looking at Sinple Sequencing, but also at Learning Design, and possibly how to join the two. In the future, IMS wants to implement progressively in Common Cartridge other standards for different goals, eg. Content Authorization Standard.
When we look at all of these standards, lots of them involve interoperability, esp. on different platforms. These do not change dramatically the quality of learning. Standards we will talk about later - Digital Repositories, Learning Design, will have more impact.
The first subject, then, is GLOBE - a network of learning repositories.
Why are we interested in learning object repositories? First, because they support resources maintained by institutions, and professors maintain their qualify through peer review and actual use. Additionally, the metadata provides valuable information about authors, subjects, etc. This enables focused queries. And the vast majority of material will be reused.
The first panel:
- Gilles Gauthier - convener at ISO SC36/WG4
- David Massart - LODE co-leader and ASPECT project manager
- Gilbert Paquette - chair (Tele-university)
Questions: are the days really over for 'one metadata for all'? What about the multiplication of profiles? Should we promote a new interchange method? Is there a need for a new standard encompassing other standards? What about the new standards, such as ISO-MLR? New projects, such as competencies?
Gilles Gauthier - Metadata for Learning Resources
The MLR is the standard development that is going on at ISO and is almost ready. It's not a big standard. Many of the people in this room work on ISO and are editors of this standard.
Gilbert asked questions. First, are the days of 'one standard' really over? My answer is 'no'. One standard is nice, if we take out the RDDF triples. Is the multiplication of profiles a problem? No. Is there a need for a new standard encompassing all approaches? Yes?
ISO/IEC JTC 1/SC 36/WG 4
ISO - International Organization for Standardization - is composed of technical committees; there are hundreds of them. The JTC is a joint technical committee with IEC - Intrnational Electtrotechnical Commission. JTC addresses standardization in the field of information technology. There afre different types of members - participating, observing and liasion.
Participating members are each representatives of countries, only one vot per country. Subgoups, called Working Groups, can also include experts nominated by participating members. There is a very clearly defined process - proposal prepatory, committee (where the voting starts), approval, and publication. The idea is to achieve consensus on a technical document.
SC36 is the committee on learning technology. SC37 is in biometrics. SC2 is the committee that produced US-ASCII and Unicode, etc. Under SC36 there are WGs (WorgGroups) including different learning technology standards. The members of SCC include national federations. Liasions include organizations like AICC, ADL, etc.
Working Groups have projects; WG4 has 9 projects, 6 related to MLR and 3 related to IMS content packaging. SCORM 2004 3rd edition became a technical report from this group.
Projects:
MLR 1 - framework. This was the hardest, defining the process for the rest.
MLR 2 - core elements. Essentially, this is Dublin Core.
MLR 3 - core application profile. This is an application profile for the core elements.
MLR 4 - technical elements
MLR 5 - educational elements
MLR 6 - availablity, distribution and intellectual property elements
CP 1 - information model
CP 2 - XML binding
CP 3 - best practice and implementation guide
Various resolutions enabled this work. One major resolution was to maintain compatibility with IEEE-LOM.
The scope is restructed to specifying, in a rule-based manner, metadata elements and their attributes for the descriotion of learning resources. This is a multipart standard, with different components. Multiple languages and multicultural requirements are supported - eg., you only use neutral identifiers., non-linguistic identifiers. All languages are equal. Also, it is intended to support multiple levels of granularity, and to support user extensions.
From a centralized resource to a distributed resource. Centralized, only one body can provide metadata. But metadata on the web can be provided by anybody. You search for the metadata, collect them, filter them, transform them, store them, explore them. So you can get from those RDF graphs MLR records. The MLR view is the centralized point of view.
The idea is to be able to specify data elements - to define what kind of data you want to have. The central element is the respurce. In the MLR all data is simple - all triples - if you want to have something complex (like a vcard) it means you want to have a resource.
Data element specifications - each has a unique identifier, each specifies exactly the possible values, each has human language equivalencies.
Data elements spec ID:
Subject: resource being described - resource ID
Content value: Fodors
Language value: eng
Resource identifier: ID
name: name
Definition: the definition
SubclassOf: other resources
Note: note
If you have a set of MLR elements, you can create an RDF binding. And if you use an XML binding, you get back something that will look exactly like an IMS-LOM record. The MLR is quite simple: it tells you how to define data elements, and how to define application profiles.
David Massart - Learning Object Exchange
I work for European Schoolnet, a network of 31 ministried of education in Europe. It is dedicated to improving learning through technologies. One of its projects is a learning resource exchange - we want all resources in Europe to be accessible.
The LRE is a service for ministries of education driven by them. With 31 countries, many languages, etc., we want 'content that travels well'. Think of music - whatever the language, it travels well. Compare with a lesosn plan in English, for example.
For the purpose of this discussion, a learning object is any resource thaqt can be used in learning. For any resource,
- you need metadata to describe it, also to assess it.
- You need the information to determine whether you want to get access to or not.
- Then once you decide you want access, you need metadata to help you use the resource.
The LRE garters metadat from ministries of eductaion, publishers, museums, teachers, etc. and makes the aggregated metadata available to LMSs, portals, etc. Various standards and protocols are involved: SOI, SPI, OAI-PMH, etc. We try to get metadata by all means, and to provide it by all means.
When you try to do that, you have to really take care of your metadata. There is a lot of negotiation, we try to agree with a provider as to what metadata must be. We control for quality and sometimes reject it. We need to correct and compile metadata, to complete the metadata records. We use an internal format:
- identifiers
- language blocks, eg. title or description in English
- indexes
We have the problem of identity. When we get a record, we need to determine whether this is a new resource, or one we already have, to update the record. Idenification of resources is key. We need to be able to uniquely identify resource - difficult because of political saspects. If you control the identity of the resource, you control the access to the resource.
IMS Learning Object Discovery and Exchange - deals with how to find repositories. There is a regitory of repositories and protocols. Learning objects can have different versions (English, French), different formats (SCORM, CC), copies, etc. This is important if you need to use an object, you need it in a format you support, in a language you read. All of these are contained in Information for Learning Object Exchange (ILOX).
Discussion
Gilbert: We have a diversity problem - it's difficult to federate, harvest, etc. And it contradicts the idea of freedom (to have one standrd), yet we want to interoperate with others. Let's suppose people are using MLR, or ILOX, how will it solve this problem.
David: I don't see diversity as a problem. We have different communities with different needs. And we are simply trying to meet their needs. You define something that is historic in your community, you start to use it. Then you interact with other communities - you negotiate. Anything that helps express the semantioc diference and express these mappings (as good).
Gilles: the difference is, you have a learning resource, it has different langauges it could be, different formasts it could be - all of those are different resources, with relations between them. Variants. Suppose you have a community that is blind. If you try to centralize them, you have more and more and more. What is a contributor: could be a person, an organization, or a service. More and more. Suppose you have a person. You have first name and last name. But thst doesn't make sense in some places. Let the community define them.
David: what you have then is a frameowrk, how to create elements and how to name them. So you have a nice framwork for identifying things. But at the end, whatever profile you use, you will start to exhange these descriptions.
Gilles: yes and no. The ISO-LOM survey from 2004 - most of gthe LOM elements people do not provide information for. If you go to MERLOT and look at the LOM record. They're mostly empty. What MLR will say, a property must have a domain, and must have a range.
Mike: I've been in meetings like this for 15 years. The purpose of metadata is discovery, wouldn;t you agree? Yes, I can find more information on Google, which uses none of that.
Gilles: As a human bneing.
Mike: as a human being. Behind Google is an incredible amount of machine leaning. Metadata has this problem. We have all thiose strucdtures, but we can't match it with the need. I still can't find what I need! How do we do it more like the way Google does, full-text indexing, etc. I understand your comments about the importance of standards, but I think we're a bit myopic.
David: it's not that you use one metadata standard rather than another. Google doesn't do such a great job when it comes to finding resources. If I want to do a specific query, I'm not sure how to express this query in Google, and I'm not sure how I will gind the resource. At the moment Google doesn't do a very good job, and we will find a better resource.
When we look at what people are soing, what we are looking at, is people who are trying to derive metadata in an automatic way. We see a text in English. We track usgae, and we see some resources travel well, and others don't. We look at the resources that do travel well, and we try to understand why. We say to providers, we are interested in these reosurces. It's a web 2.0 approach - we look at tags, bookmarks, keywords, descriptors, etc. We see uses that we didn't forsee.
The key of the ILOX approach is that it's a container. We try to encapsulate metadat created by providers, but also taxonomies, ratings, etc., generated by users. We don't make any assumption on what is a useful metadata or not.
Gilles: I also agree, Google, for human beings, you will find something, but not much use for machines. I would liek to see, just another way. You have a resource, and you have an identifier. You can have as many identifiers as you want, you can go to the international identifiers union and get a unique ID for free. I would like to ask Google, this record I want to have, and then there is a graph of related resources, this resource is the blind, etc. Google could do a better job presenting links.
Comment: Good, MLR aligns with the semantic web, which is where it has to go. Now, as an abstract model, it sounds a lot like the Dublic Core abstract with the semantic web. All of those seems to align one-for-one seem to align with those movements. How does MLR differ from them?
Gilles: Global ID is the same. What we add is a ay to specify an applucation profile to get LOM-like records. Also, the way this is multilingual, using a neutral identifer. All those data elements will be on an international standard.
Comment: when you talk about the URIs, are you just saying they are opaque, that they don't carry semantic data?
Gilles: they could. It's just the way you create your URIs. They have structure. You have rule sets. This is very precise, to speify the domain, the range. This is a namepace for ISO. ISO has a whole structure. We will use that.
David: The problem being if people assign different unique identifiers to the smae resource.
Gilles: You won't have that. You will have one.
Gilbert: you've seen two major initiatives to try to reconcile free and centralized descriptions of resources. These two initiatives can probably link together. ILOX can use MLR. And what W3C is puting out.
Diny: we've gone throught he same process with GEM, and 15 million resources with metadata that needs to be massaged. We have transition that store to an RDF store. So we live in both worlds now - the XML and the RDF world. The blending is not only attainable, but we're doing it.
