Explaining Linked Data to Your Pro Vice Chancellor

At the JISCEXPO Programme meeting today I led a session on ‘Explaining linked data to your Pro Vice Chancellor’, and this post is a summary of that session. The attendees were: myself (Adrian Stevenson), Rob Hawton, Alex Dutton, and Zeth, with later contributions from Chris Gutteridge.

It seemed clear to us that this is really about focussing on institutional administrative data, as it’s probably harder to sell the idea of providing research data in linked data form to the Pro VC. Linked data probably doesn’t allow you to do things that couldn’t do by other means, but it is easier than other approaches in the long run, once you’ve got your linked data available. Linked Data can be of value without having to be open:

“Southampton’s data is used internally. You could draw a ring around the data and say ‘that’s closed’, and it would still have the same value.”

== Benefits ==

Quantifying the value of linked data efficiencies can be tricky, but providing open data allows quicker development of tools, as the data the tools hook into already exist and are standardised.

== Strategies ==

Don’t mention the term ‘linked data’ to the Pro VC, or get into discussing the technology. It’s about the outcomes and the solutions, not the technologies. Getting ‘Champions’ who have the ear of the Pro VC will help.  Some enticing prototype example mash-up demonstrators that help sell the idea are also important. Also, pointing out that other universities are deploying and using linked open data to their advantage may help. Your University will want to be part of the club.

Making it easy for others to supply data that can be utilised as part of linked data efforts is important. This can be via Google spreadsheets, or e-mailing spreadsheets for example. You need to offload the difficult jobs to the people who are motivated and know what they’re doing.

It will also help to sell the idea to other potential consumers, such as the libraries, and other data providers. Possibly sell on the idea of the “increasing prominence of holdings” for libraries. This helps bring attention and re-use.

It’s worth emphasising that linked data simplifies the Freedom of Infomataion (FOI) process.  We can say “yes, we’ve already published that FOI data”. You have a responsibility to publish this data if asked via FOI anyway. This is an example of a Sheer curation approach.

Linked data may provide decreased bureaucracy. There’s no need to ask other parts of the University for their data, wasting their time, if it’s already published centrally. Examples here are estates, HR, library, student statistics.

== Targets ==

Some possible targets are: saving money, bringing in new business, funding, students.

The potential for increased business intelligence is a great sell, and Linked Data can provide the means to do this. Again, you need to sell a solution to a problem, not a technology. The University ‘implementation’ managers need to be involved and brought on board as well as the as the Pro VC.

It can be a problem that some institutions adopt a ‘best of breed’ policy with technology. Linked data doesn’t fit too well with this. However, it’s worth noting that Linked Data doesn’t need to change the user experience.

A lot of the arguments being made here don’t just apply to linked data. Much is about issues such as opening access to data generally. It was noted that there have been many efforts from JISC to solve the institutional data silo problem.

If we were setting a new University up from scratch, going for Linked Data from the start would be a realistic option, but it’s always hard to change currently embedded practice. Universities having Chief Technology Officers would help here, or perhaps a PVC for Technology?

Putting the Case for Linked Data

This is a summary of a break-out group discussion at the JISC Expo Programme meeting, July 2011, looking at ‘Skills required for Linked Data’.

We started off by thinking about the first steps when deciding to create Linked Data. We took a step back from the skills required and thought more about the understanding and the basic need and the importance of putting the case for Linked Data (or otherwise).

Do you have suitable data

Firstly, do you have data that is suitable to output as Linked Data. This comes down to the question: what is suitable data? It would be useful to provide more advise in this area.

Is Linked Data worth doing?

Why do you want Linked Data? Maybe you are producing data that others will find interesting and link into? If you give your data identifiers, others can link into it. But is Linked Data the right approach? Is what you really want open data more than Linked Data? Or just APIs into the data? Sometimes a simpler solution may give you the benefits that you are after.

Maybe for some organisations approaches other than Linked Data are appropriate, or are a way to start off – maybe just something as simple as outputting CSV. You need to think about what is appropriate for you.  By putting your data out in a more low-barrier way, you maybe able to find out more about who might use your data. However, there is an argument that it is very early days for Linked Data, and low levels of use right now may not reflect the potential for the data and how it is used in the future.

Are you the authority on the data? Is someone else the authority? Do you want to link into their stuff? These are the sorts of questions you need to be thinking about.

The group agreed that use cases would be useful here. They could act as a hook to bring people in. Maybe there should be somewhere to go to look up use cases – people can get a better idea of how they (their users) would benefit from creating Linked Data, they can see what others have done and compare their situation.

We talked around the issues involved in making a case for Linked Data. It would be useful if there was more information for people on how it can bring things to the fore. For example, we talked about set of photographs – a single photograph might be seen in a new context, with new connections made that can help to explain it and what it signifies.

What next?

There does appear to be a move of Linked Data a ‘clique’ into the mainstream – this should make it easier to understand and engage with. There are more tutorials, more support, more understanding. New tools will be developed that will make the process easier.

You need to think about different skills – data modelling and data transformation are very different things. We agreed that development is not always top down. Developers can be very self-motivated, in an environment where a continual learning process is often required. It may be that organisations will start to ask for skills around Linked Data when hiring people.

We felt that there is still a need for more support and more tutorials. We should move towards a critical mass, where questions raised are being answered and developers have more of a sense that there is help out there and they will get those answers. It can really help talking to other developers, so providing opportunities for this is important. The JISC Expo projects were tasked with providing documentation – explaining what they have done clearly to help others. We felt that these projects have helped to progress the Linked Data agenda and that it is an important encouraging people to acquire these skills to require processes and results to be written up.

Realistically, for many people, expertise needs to be brought in. Most organisations do not have resources to call upon. Often this is going to be cheaper than up-skilling – a steep learning curve can take weeks or months to negotiate whereas someone expert in this domain could do the work in just a few days. We talked about a role for (JISC) data centres in contributing to this kind of thing. However, we did acknowledge the important contribution that conferences, workshops and other events play in getting people familiar with Linked Data from a range of perspectives (as users of the data as well as providers). It can be useful to have tutorials that address your particular domain – data that you are familiar with.   Maybe we need a combination of approaches – it depends where you are starting from and what you want to know.  But for many people, the need to understand why Linked Data is useful and worth doing is an essential starting point.

We saw the value in having someone involved who is outward facing – otherwise there is a danger of a gap between the requirements of people using your data and what you are doing. There is a danger of going off in the wrong direction.

We concluded that for many, Linked Data is still a big hill to climb. People do still need hand-ups. We also agreed that Linked Data will get good press if there are products that people can understand – they need to see the benefits.

As maybe there is still an element of self-doubt about Linked Data, it is essential not just to output the data but to raise its profile, to advocate what you have done and why. Enthusiasm can start small but it can quickly spread out.

Finally, we agreed that people don’t always know where products are built around Linked Data. So, you may not realise how it is benefitting you. We need to explain what we have done as well as providing the attractive interface/product and we need to relate it to what people are familiar with.








Describing the “things”: the RDF terms used (part 1)

In previous posts, I described:

  • the model of the “world” on which we’re basing the Archives Hub RDF data: the types of “thing” being described, and (some of) the relationships between them (1, 2, 3); and
  • the patterns for URIs to be assigned to the individual “things”

In this post and the next one, I’ll outline the RDF vocabularies we’re using to describe those “things”. This post covers some of the considerations in choosing the vocabularies and some of the “patterns” we’ve used in deploying them; the next lists the properties and classes you can expect to find in the LOCAH data.

Using existing RDF vocabularies

As far as possible, we’ve tried to make use of existing, deployed RDF vocabularies. These include:

Those distinctions between which vocabulary “describes” what are somewhat rough, particularly taking into account that the “directionality” of properties in RDF is somewhat arbitrary: a triple using the dcterms:creator property to link a created work to an agent is as much “about” the agent as it is “about” the thing created.

However, where we’ve seen a need to express a notion that is not well addressed by an existing vocabulary, we have defined the additional classes and properties required and provided URIs for them as a small “local” LOCAH RDF vocabulary. At this point in time, I consider most of these terms something of a “work in progress”, and likely to be revised (or even dropped completely) before the end of the project. But I suspect some will remain – which, given the bounded timescale of the project, leaves questions about the longer term management of such vocabularies.

Discovering Appropriate Vocabularies

Most of my knowledge of existing RDF vocabularies has come from lurking on good old-fashioned mailing lists, particularly the W3C Semantic Web Interest Group list and the Linked Open Data list. I don’t read every posting by any means, and the signal-to-noise ratio can be variable, but for me they remain an excellent source of information with a knowledgeable and active contributing community (and the archives are a great repository.)

In similar territory, Semantic Stackoverflow provides a “question-and-answer”-style service, though it tends to have a fairly technical focus.

Another useful source is to look at actual linked data datasets, particularly those which are in a similar “domain” to the one you’re working in and cover similar resource types, and check out what vocabularies they are using (and how they are using them). In the library/bibliographic domain in particular, there has been a fairly steady stream of linked data datasets appearing over the last couple of years, so there’s quite a bit to go on, rather less so for the archives case. For a few pointers, see e.g. this review post by Ed Summers (itself already nearly a year old).

There are some services which aim to provide disclosure/discovery services based on aggregations of information about vocabularies and their constituent terms, sometimes called “metadata registries” or “metadata schema registries”. I’ve had mixed experiences of using these services: in some cases the content is not current; in others the coverage is intentionally tailored to the requirements of a particular community, so the challenge becomes one of finding a registry whose coverage matches the task at hand. One service (with quite general coverage) which I have occasionally found useful is Schemapedia, a project by Ian Davis of Talis; it provides “vocabulary”-level descriptions, rather than descriptions of individual “terms” but it includes some examples of actual terms: see, e.g. the entry for the Biographical Vocabulary.

There are a number of services which provide search functions across aggregations of data gathered from the linked data Web/Semantic Web. Sindice crawls and aggregates a huge range of RDF data and provides a “Google”-like search across that aggregation. (I’ve also found navigating such an aggregation helpful in thinking about various aspects of linked data: the sig.ma browser highlights the consequences of merging data from multiple sources, and related issues of provenance, attribution and trust, for example).

Finally, at the risk of stating the obvious, plain old Web search engines can still be a useful entry point.

Having said all this, I admit that the discovery of RDF vocabularies is still something of a challenge, and I continue to come across useful things I’d missed. And having found something potentially useful often raises further questions: Is the vocabulary stable or still being developed? Is it described following “modern” good practice for RDF vocabularies? Is it being managed/curated? By an individual/institution/community? Does it have the support of a community of users? Particularly if the intention is for a dataset to have some longevity, these may be significant considerations.

Patterns for using RDF Vocabularies

While discovering RDF vocabularies capable of expressing the information you want to represent is a first step, it often raises issues of exactly how those vocabularies might best be deployed, or of choosing between several possible alternative solutions.

Leigh Dodds and Ian Davis of Talis have authored a booklet Linked Data Patterns which tries to address some of these challenges, by gathering together some common “patterns” of use, based on existing practice by linked data implementers – though perhaps inevitably at this stage, some aspects of that practice are something of a “moving target” as new challenges are identified and practice evolves to address them. (See, for example, a recent debate on the Linked Open Data mailing list covering the question of expectations for what the object of an rdfs:seeAlso triple might/should dereference to.)

I continue to find the reflections of linked data practitioners an excellent source, particularly those working in domains close to those I’m interested in. I regularly find myself referring to the series of posts by Jeni Tennison on creating linked data. In this context, the fifth post on “Finishing Touches” is particularly relevant, and in large part prompts my next couple of points.


One of the principles I’ve tried to adhere to, following the guidance by Jeni is that each resource we expose should have a human-readable label, provided using the rdfs:label property, and as far as possible that label should function as a useful “stand-alone” name for the thing.

In some cases this is a straightforward matter of using some text content node in the EAD XML document as an RDF literal. In other cases, a single element in the EAD document is mapped to a number of distinct resources in our model. In these cases, the transformation process typically prefixes or suffixes the source text to generate labels for the various different things. Perhaps unsurprisingly, this sometimes leads to some slightly “artificial” or “stilted” results, so it’s something we may need to refine.

Also, and perhaps more problematically, as I’ve noted in a previous post, the practice of archival description has traditionally relied heavily on a “multi-level description” approach which results in the presentation of resource descriptions “in the context of” the descriptions of other related resources. So it is common to find individual items within a collection labelled simply as something like “Letter”, on the basis that the reader of the finding aid will glean further information from the fact that the description of the item is presented within a context provided by a list of other “sibling” items, all “children” of a “parent” aggregation of some form. Currently our mapping generates the rdfs:label of an item using only the label (EAD unititle element) of that item in the EAD document, with the result that we may indeed end up with many individual resources labelled “Letter” (though of course the description will also include other properties derived from other EAD data and links to “parent” resources). An alternative might be to try to generate a label by “qualifying” the item unittitle, say, by prefixing it with the label of a “parent” resource – though I suspect in practice this would generate some somewhat unwieldy results.

Where the source data makes it seem reasonable to express it, I’ve also indicated the use of a “preferred label”, using the skos:prefLabel property. I’m conscious here of the need to be careful: the SKOS specification includes a number of “integrity conditions”, rules which data using the SKOS vocabulary should follow. Amongst them is the requirement that

A resource has no more than one value of skos:prefLabel per language tag.

The important thing to remember is that this is intended to apply in an “open world” context, not simply as a condition scoped to a particular “document”. The EAD to RDF transform process is performed on a document-by-document basis. Within the Hub dataset, it is quite common that for a single resource, labels for that resource are generated from the content of multiple EAD documents. While in theory naming within the set of EAD documents should be consistent, in practice, the use of variants of names is widespread in our data – the names of archival repositories is one example. Generating an skos:prefLabel triple for each variant would result in a conflict with the integrity condition once the data was merged in the triple store.

Bearing in mind that the “open world” extends beyond the boundaries of our own dataset, the same considerations apply in the case where we are exposing URIs for resources for which other parties already expose descriptions, including an skos:prefLabel triple, and we can’t guarantee that the names in our data correspond to those provided by that source.


Another issue to consider is that referred to by Leigh and Ian in their “Materialize Inferences” pattern, and by Jeni Tennison in her discussion of “Derivable Data”. One of the strengths of using the RDF model is that it is supported by a formal semantics, a framework for reasoning with data, i.e. given some set of data, it is often possible to apply some formalised set of rules to infer or derive additional triples. However, it should not be assumed that all consumers of the data will have access to the tools which support such reasoning, so it may be more appropriate for a data provider like LOCAH to explicitly include at least some of those “derivable” triples in the data we provide.

For a simple example of what I mean, the Friend of a Friend (FOAF) vocabulary provides a property called foaf:name (“A name for some thing.”). As part of their description of that property, the FOAF vocabulary owners provide the triple:

foaf:name rdfs:subPropertyOf rdfs:label .

The RDFS property rdfs:subPropertyOf is one of those properties which is associated with a set of rules. What those rules say is that, for any two properties linked by an rdfs:subPropertyOf relation, two resources related by the first property are also related by the second. So each time I find a triple using foaf:name as a predicate, I can infer (deduce, derive) a second triple using the rdfs:label predicate, e.g. if I find

<http://example.org/id/person/p123> foaf:name “Ernest Henry Shackleton” .

then I can conclude

<http://example.org/id/person/p123> rdfs:label “Ernest Henry Shackleton” .

However, to reach that conclusion, my application needs (a) knowledge of the general rdfs:subPropertyOf inference rule, and (b) knowledge that foaf:name is a subproperty of rdfs:label – and (c) the processing capability to apply that rule!

By providing – “materializing” – both those triples in our source data, we relieve the consuming application of that responsibility – though that benefit comes at the cost of increasing the size of the descriptions we provide.

This tactic can be particularly useful, I think, for properties which are subproperties of “generic” vocabularies like the RDF Schema vocabulary or the Dublin Core vocabularies. Sometimes generic linked data tools have some “built-in knowledge” of, and/or specific behaviour associated with, some of these vocabularies (e.g. to obtain literal names/labels/titles for display to human readers). It may be perfectly reasonable to use a triple with some more specialised subproperty in our data to indicate some specific relationship, but where appropriate it is also helpful to “materialize” the triple using the more generic property as well, so that an application looking for RDF Schema or DC properties can easily access that data.

Extending that slightly, Jeni suggests a “rule of thumb” that “if the result of the reasoning involves a resource from another vocabulary, then we should include it”.

The subproperty case is just one example: the inference of resource type based on rdfs:range and rdfs:domain is another case in point. In the LOCAH data, we’ve tried to provide fairly “generous” type data (e.g. including “super-classes”) where possible – again, on the grounds that such information is a commonly used “hook” in user queries (“Select resources of type T where [some other criteria]”).

The “cost” of this approach is that the dataset and the individual “bounded descriptions” served are larger – so there is a “trade-off” here which we may want to monitor and reconsider once we see how the data is being used.


As I mentioned earlier, we extended our very initial draft model to include a notion of “event”. Currently, the application of this approach in our data is quite limited: it is applied to the “creation”/”origination” of the archival resources, and to the birth, death and “periods of activity” (floruit) of individuals. What we do is similar to the approach sketched by Ben O’Steen in his processing of the British Library’s British National Bibliography data – though with a little more complexity as we make use of event ontologies which model time periods as resources, rather than as literals.

This is probably best illustrated by means of an example. Given a person with birth date of 1901 and death date of 1985, we generate an RDF graph like the following:

RDF Graph of Life Events Data

RDF Graph of Life Events Data

(The image links through to a larger version)

The time interval nodes at the right-hand side are reference.data.gov.uk URIs for years, like http://reference.data.gov.uk/id/year/1901

What I haven’t illustrated on that diagram is that I’ve also included some data using the CIDOC CRM ontology – actually using the Erlangen CRM vocabulary. I’m feeling my way a bit with this, so it is somewhat partial/experimental at the moment, but I hope to refine/extend it in the future.

The point I wanted to highlight is that we’ve made use of multiple “overlapping” vocabularies here – again on the grounds that it may be useful to provide that flexibility to consumers of the data querying using a specific vocabulary. As above, this is a “trade-off” which we may want to monitor and reconsider in the future.


I’ve tried to cover here some of the issues around our choices of RDF vocabularies and how we’ve deployed them. The next post will summarise the actual terms used.

Assessing Linked Data

For me, the journey from an understanding of modelling data, and creating our own models for the Hub and Copac, to being able to understand the processes and decisions involved in creating XML RDF has been challenging. It has raised one question that often applies when dealing with something quite technical: how much should a manager (in my case an archivist managing an online archive service) be expected to understand the ‘technical’ aspects of something?  This is a question I have spoken about and written about before; mainly in terms of what archivists (or other information professionals) in the digital age need to know in order to understand the implications of choices around things like data structure and software systems.  In the case of Linked Data, I am still not sure how much I need to know about the detail of Linked Data, the RDF model, the use of RDF XML, the benefits of other output fomats, the application of stylesheets, etc. I have been thinking about how hard it is to create Linked Data – I have had a few enquiries from colleagues who are interested in doing the same sort of thing already and I want to be able to offer useful advice.

One thing that occurs to me is that it is reasonable to acknowledge that Linked Data does involves programming skills, and therefore it is not so dissimilar from structuring and outputting your data through a traditional relational database, for example, where you would expect that specialist skills are needed. But in either case there is the the same need for a manager to understand what the system offers and be able to offer the best service to researchers. I think what is important from the point of view of the manager is to be involved with the decision-making process and understand the implications of Linked Data; you need to know what you are saying about your own data. I am not sure that this requires a thorough knowledge of RDF and certainly I only have a rudimentary knowledge of stylesheets (and no knowledge of programming).

Service managers or administrators are not expected to understand systems from an in-depth technical point of view. But in fact, I think that one of the advantages of RDF model is that it is easier to get a sense of what is going on in terms of data processing than typically occurs with a database management system. After about six months of learning about Linked Data and RDF (I estimate that this translates into about one month of fairly intense learning), I can look at the stylesheet that we have for the Archives Hub, to transform the EAD data into RDF XML, and I can look at an RDF document representing the entities that we are describing, and I have a reasonably good overall sense of what it all means, which helps me with my main role: to understand the outputs and the potential benefits of Linked Data. For Locah, we’ve used XSLT, but there is no requirement for this, and maybe one of the challenges of outputting Linked Data is that there are a number of options in terms of translating your RDF model into an output.

There is no doubt that choices made now about how we model the data will have implications for what users can do with it, and some choices may limit future potential more than others. For example, which ‘things’ do we choose to represent? Should we have a conceptualisation of a person as an entity represented in the description, and to link this to a conceptualisation of the person? Which information should we provide as URIs and which as literals? I’m only gradually coming to understand the implications of these decisions, as we start to explore the potential of the data. Of course, this is always true, whatever data, structures and systems we are working with. This brings me to another point that I think is probably particularly relevant for Locah: we are doing this work at a time when we are very much early adopters. Whilst the classic Linked Data diagram may give the impression that the world has embraced Linked Data, the reality is that it is still very much at a hand-crafted level: we have not had tools available to us to aid us in this work, and in the case of EAD, there has been very little activity up till now. It is therefore difficult to judge how feasible it might be to output RDF in the future, as it is likely that more tools will be developed, and there will be greater awareness and skills built up around the whole Semantic Web. However, I wonder if we are currently still at that difficult point where we need to build the momentum of the Linked Data movement, but it is still very unfamiliar and poorly understood by many data providers?

Many Linked Data evangelists claim that Linked Data is ‘easy’. I’m not sure that it is necessarily easy, and I don’t think that it’s very helpful to say that it is easy. Easy compared to what? Easy for whom? It’s easy if you know how, if you have the requisite skills and experience, but we need to persuade people who don’t yet know how that it is worth doing, and provide a realistic assessment of the skills that are required. I suppose the question of how easy it is does rest in large part on the data you are working with as well. Archival finding aids are quite challenging. As Mark Matienzo, archivist at Yale University, states in his presentation on Linked Data and Archival Description: “Archival description is inherently multi-level and relational” and “EAD is both too flexible and too unforgiving” to be Linked Data friendly…and database-friendly for that matter. Also, ISAD(G) recommends the non-repetition of information and archival description generally contains implicit information. I suppose Linked Data might help provide the opportunity and impetus to move towards a more Web-friendly way of describing archives, if it does become more widely used.

At present, I can’t help thinking that if archive repositories and libraries would like to output their data as Linked Data, many of them will struggle, and I would have thought it might be similar for other types of data providers. I do think that expertise is required, and time needs to be invested in understanding some key aspects of Linked Data. On the other hand, this is the case whenever you are looking at creating effective means to output structured (but often inconsistent) data. However, I think that it makes good sense for the Archives Hub and Copac to do this work, as it is on behalf of our contributors, so it effectively will allow these repositories and libraries to output Linked Data.  In other words, it may be that for Linked Data to really take hold, it will benefit from this kind of aggregated set-up, where skills and resources can be pooled. At present, I’m inclined to think that it is worth the investment of time and resources by our Locah team because it is benefitting a large number of data providers. I think it will be important for us to convey to our contributors, and indeed to other archivists and librarians, what we are doing and why, what the implications are and what the benefits may be. I have already had contact with two people, one representing another aggregation of content, interested in benefitting from our work. This is really important, because it potentially makes the investment more worthwhile.

We are in a fortunate position with the Locah project because we are part of a JISC-funded innovations project, with a team of people with a variety of skills, and we have support from Talis, who have significant experience of Linked Data.  If we can work on behalf of our community, then I feel that the time invested may be worthwhile. For the second half of our year-long project we will want to explore the benefits more thoroughly – we will be looking at the crucial issues of creating links to other data, which is really Linked Data’s key selling point, and we will be developing a prototype to show some potential benefits for researchers.

(With thanks to Pete and Ade for their contributions to this blog post).

Modelling Copac data

With the Archives Hub data well under way, it was time to start looking at the Copac data.  The first decision to be made was which version of Copac data to use – consolidated or unconsolidated.  As part of the process of adding records to Copac they are de-duplicated, allowing different institutions’ records for the same item to be presented as one record, instead of several.  For more info on Copac de-duplication, see this blog post.

So our first question was: deal with the individual records from each library, or with the consolidated records created for Copac?  This made us think about the nature of what we were describing. The unconsolidated records (generally!) relate to the actual, physical ‘thing’ – what in FRBR would be the ‘item’.

The consolidated records are closer to (but by no means a perfect example of) the FRBR manifestation.  That is to say, they are describing different physical instances of the same theoretical work; in linked data terms, they are ‘same as’.  They aren’t perfect manifestation level records, as there may be other records on Copac for the same manifestation which haven’t been consolidated due to cataloguing differences.  At Copac, we err on the side of caution, and would rather have this happen, than have records which aren’t the same consolidated into the same record.

So we could do our mapping and our transformations at unconsolidated level, and then use ‘same as’ to link together the descriptions that would later be consolidated in Copac.  But as we’re accepting Copac’s judgement that they are describing the same set of items, why not save ourselves that trouble, and work from the consolidated description?  We can then hang the individual bibliographic records off this central unit of description.

This means that all of the information provided by the different libraries is related to the same unit of description.  The bibliographic records that go together to make up a consolidated Copac record may not contain all of the same information, but they won’t contain any contradictory information.  Thus two records which are the same in all details except date of publication (say 1983 in one, 1984 in the other) will not consolidate, but records which are the same in all details except that one contains a subject where the other does not, will consolidate.

In fact, subjects are one of the things (along with notes) that don’t affect consolidation at all.  We will combine all of the subjects that come in individual descriptions, so that a consolidated record might end up with the subjects:


Management — theory

Management (theoretical)

Business & management

We will leave these in the linked data description for the same reason they are left in the Copac description – while such similar terms may seem superfluous, they actually increase discoverability, by providing multiple access points.  They will link into the central ‘unit of description’, rather than the individual bibliographic records.

Once we’d decided on this central unit of description (name TBD, but likely to be ‘Copac record’ or something boringly similar), other aspects of the description started to fall into place.  Some of these were straightforward – publication date, for instance, is fairly obviously a literal – while others took more thought and discussion.

Among the more complicated issues was that of creator.  We are working with MODS data, which has come from MARC data, and MARC allows you to have only one ‘creator’.  This creator sits in the 100s as the main access point, and all other contributors (including co-authors!) are relegated to the 700s, where they become what we have decided to call ‘other person associated with this unit of description’.  Not very snappy, but hopefully fairly accurate.  In theory, the role that this person has in the creation of the item should be reflected in a MARC indicator, but in practise this is not often included in descriptions.  Where it is indicated that a person (or a corporate body) is an editor, contributor, translator, illustrator etc, we can build these into the modelling; where not, they will have to be satisfied with the vague title of ‘associated person’.

This will work for most situations, but it does still leave room for error.  Where a person is named in the 700s with no indicator of role, it is possible that they are a person who was associated with one particular item, rather than the manifestation – a former owner or bookseller, for example.  While we do want to present this information, which works as another access point, and may be of interest to users, we have the problem that this information should really be associated with the item, not our quasi-manifestation.  This information only concerns one specific physical item, as described in one of the individual bibliographic records. Should it really have a link to our central unit of description?  If not, where do we link it to?  Our entries for individual bib records describe only the records themselves, not a physical real-world item.  It’s an interesting point, and one we’ll be dicussing more as the project goes on.

We’re continuing to work with Copac data, and will discuss other issues here as they arise.

Creating Linked Data: more reflections from the coal face

This post is to highlight some of the barriers and challenges to the creation of Linked Data.  This is a personal reflection, trying to be honest about the challenges as I have found them and the learning experience, which is inevitably a personal thing depending upon your own background, experience and ways of thinking and working. However, I think it also reflects some of the general challenges as we have come across them.


It comes as no surprise that I have found the terminology somewhat confusing, and it has sometimes led me astray. Only this week Bethan and I were getting tangled up in a conversation about ‘things’  within the data model. We spent a while talking about how having a ‘Hub conceptualisation’ and a ‘thing-in-its-own-right conceptualisation’ of an entity would allow for more clarity. With ‘thing’, ‘concept’, ‘label’, ‘property’, ‘value’, ‘predicate’, ‘information resources’, ‘non-information resources’ etc. – there is quite a bit of room for misinterpretation in communication. I have looked at definitions, but these can actually sometimes hinder rather than help. I think that an attempt at a definitive glossary for Linked Data would help enormously.


For me, it has taken a while to really get into the Linked Data way of thinking. I have actually kept a kind of diary of my thoughts over the last 2-3 months, and when I look back now at my earliest attempts at understanding how to model the data, they certainly show a pretty steep learning curve. I started, for example, by being unsure about whether we were wanting to provide information on the ‘creator’ of the archive or the archive itself and what sort of relationships between ‘things’ to include. I don’t think this is surprising, as the power of RDF is that it can be used to model anything – it doesn’t help you by giving you a limited scope or particular rules to start with (which is, of course, generally a good thing).

Archival descriptions

I listened to a number of audio tutorials, read a number of reports, blogs, etc., and learnt a great deal from these, but I still found the lack of examples within my own particular domain to be a barrier. Talis provide a very excellent tutorial that you can sit and listen to, but the real-world example is for a whiskey distillery. It somehow seems a long way away from an archival description! So, I would definitely say this lack of information for my domain was a barrier. But, of course, for others who want to output their finding aids as Linked Data in the future, we should start to see models developing that they can use, with examples and information to help (Locah, we hope, being one source of help).

Expertise and experience

The Locah team has a variety of expertise and experience, but it is undoubtedly true to say that I would be struggling a great deal more than I have done if we had not had the input of Pete Johnston from Eduserv, who has been very much involved in the EAD modelling. Whilst it is important (and pleasant) to give credit where it’s due,  the real point here is actually that I think a certain level of expertise is important, to model data and output RDF. I have experience as an archivist and understand EAD and metadata, Pete also has experience of working with archival descriptions, and also substantial experience of metadata standards and issues around the Semantic Web and technical interoperability. We also have Bethan Ruddock working with us, who now has 18 months experience of working with EAD descriptions, and is a trained librarian. That is just the core team looking at the archival data modelling.  In addition, the expertise of UKOLN will come into play with other aspects of the project.

I find it hard to see how this sort of work could currently be done by a team with substantially less experience in these sorts of areas. However, it is important to state that we will also be working with Talis, who have a great deal of expertise in Linked Data. They are providing access to their own Triple Store and other benefits that we can take advantage of. Others thinking of outputting Linked Data could look to involve companies like Talis more heavily, thus taking advantage of their expertise and requiring less in-house expertise.

The benefits of data modelling

One of the areas that I spent most time trying to find good tuturials about was data modelling. I may have missed some things that would have been very useful, but as it is I found that there simply wasn’t enough helpful information about how to create a data model. This would have saved me quite a bit of time because I think the data model is so central to what we are doing and provides such an effective way to visualise the entities and relationships between them. I think this was partly a case of examples being too simplistic, and partly a lack of data models that used catalogue data – not necessarily archival finding aids, but at least something similar.

The data

I think that we are going to find challenges around the actual content. There are numerous examples of inconsistencies, such as where the ‘creator’ is ‘Joe Bloggs and others’ rather than just a name, or where the access points do not have rules or a source associated with them. I’ve just found some descriptions where the content for the ‘extent’ should acutally really be in the ‘scope’. Some descriptions have rather unsatisfactory references, some do not include the language field, a few do not even include the creator field. For some fields we will just be outputting literal values, but for others consistency would help a great deal with the creation of RDF, particularly when thinking about the vocabulary (or predicate) that we use to define the relationship between a subject and object.  This is the challenge of creating Linked Data for descriptions that have been created by 200 different institutions over several decades and by 100s of different people. We’ll have to see how it goes!

The issue of access points

Within EAD there are access points, or index terms, associated with the description. These are most commonly subject, name and place. We’ve found that establishing the nature of the relationship between the unit of description and the access point is not easy. It looks like the relationship is going to be something very unspecific, such as ‘associatedWith’. I’m not sure yet whether this has any implications…


For me, after a few weeks away from thinking about Locah and Linked Data, getting back into the whole mindset actually takes about an hour and a nice cup of tea. In other words, the mindset I require to think about Linked Data currently feels separate from my normal working mindset. I think this is because LD requires something different. This in itself makes it quite challenging. It doesn’t fall naturally into what we do in the Hub and how we think about metadata.

However, the very big plus with this different kind of thinking is that really by definition it puts what the user is interested in at the forefront of your thinking. Well, maybe I should qualify that: I believe it puts what the user is interested in at the forefront. This is because we understand that users of archives are usually primarily interested in individuals, families, organisations, subjects and places. What they want is information on Sir Ernest Shackleton, Barbara Castle, Victorian theatre, town planning, a local business, a scientific organisation, the history of Manchester the industry of Sheffield,  or anything else. They don’t tend to know that they want to access a particular archive. Or if they do, it is often due to an assumption that there is ‘an archive’ on the person or organisation that they are researching. Even if there is an archive, there may may be a misplaced assumption that this archive is pretty much all the stuff about that entity. Furthermore, there are going to be many many researchers out there who will not be aware of archives and how to access them.  Linked Data provides a way to link archives into…well, into just about anything else.

Making sense of modelling EAD

Last week Pete Johnston, Bethan Ruddock and I got together and shut ourselves in a room for 5 hours with a whiteboard, flipchart and with our thinking caps on. Pete has already posted some thoughts about architecture and workflows following this meeting. I thought I would share some more informal thoughts of my own – from the perspective of an archivist and someone gradually getting to grips with Linked Data and RDF modelling.

Now that I understand a bit more about RDF, I can see where some of my misunderstandings were leading me astray. Firstly, it took me quite a while to get away from the idea of modelling the EAD record, rather than the actual data. This might seem obvious to those conversant in Linked Data, but I’ve been dealing with records as the unit of information for the last 20 years. With Linked Data you have to get away from this and think about the actual concepts within the data. The record (the EAD description in this case) exists as an entity along with everything else, but it can be misleading to take it as the starting point for data modelling.

I found actually getting a ‘starting point’ a bit difficult. I think this is because everything can be a starting point, and also because I kept going back to thinking of something like <http://archiveshub.ac.uk/search/record.html?id=gb15sirernesthenryshackleton> as the starting point (the record itself). I then moved away from this and started thinking about the archival creator as a central concept. I knew that in RDF this person or organisation would be a subject. I also knew that this subject would need a URI and that we might want to tell people about stuff related to this subject, but I struggled with how we would provide URIs for subjects like this, and also how we would link the creator as subject to things like the index term subjects.

After a quick chat with Pete Johnston I started to understand the real role of URIs within Linked Data. We are probably going to create URIs ourselves for things (concepts) within the Hub. So, we might create a URI for every archival creator, and a URI for every repository, etc. We agreed that we needed to model the data within our world before looking too much at linking to data outside of it.  Whilst I had listened to, and read a good deal of literature on Linked Data, I somehow hadn’t quite got the idea that you might create URIs yourself for your own concepts and that these would be documents in their own right, so then you can link to these URIs within your statements and you can include whatever information you think will be useful within these documents.

For example, we were using Sir Ernest Henry Shackleton as a sample record (the famous Antarctic Explorer). He would have a URI – something like archiveshub.ac.uk/id/person/sirernesthenryshackleton. By providing him with a URI we can then create triples (statements) that include this URI. For example:

archiveshub.ac.uk/id/person/sirernesthenryshackleton ‘created’ http://archiveshub.ac.uk/search/record.html?id=gb15sirernesthenryshackleton.

We can then decide what information we will put in this document that identifies Sir Ernest, so that when researchers look up the URI, they get useful information. We can include links to external locations and we can look at using the ‘sameAs’ relationship to link to other representations of the same person.

Some URIs are fairly straightforward. We will create URIs for archival levels, and then these can in theory be used by others who want to identify levels within the data. For something like language, we will probably use URIs that are already available.

It is useful within data modelling to distinguish the real from the conceptual. So, going back to Sir Ernest, he is a flesh and blood person, and he can also be represented as a concept. If we are thinking about subjects used as index terms within the data, you might have ‘Exploration’ as a subject. We want Sir Ernest, the man described within our description, to be associated with this subject, so we can do this by making him into a concept, and giving that concept a URI. We can then link that to a literal value – his name. In our meeting we discussed one of the advantages of conceptual agents as being that we can distinguish between the person or organisation in its entirety and the person or organisation within this particular context. Archives often only represent a small part of someone’s life or an organisation’s activities, so it is helpful to talk about ‘Sir Ernest Shackleton’ as the explorer and leader of the British National Antarctic Expedition of 1907-1909.

So, we are now starting to move towards a model where we have URIs for a number of key concepts within the Hub. Our intention is to limit the number of concepts that we create URIs for, at least at this stage. We will also simplify some areas with the EAD modelling that we can then open up for investigation later on. For example, it would be good to look at version control and how we might filter changes to Hub descriptions through to the RDF XML, but we think that initially it is a good idea to create Linked Data from our basic model so that we can get feedback and also benefit from the learning process.

The main text heavy field that we are planning to create URIs for at this stage is the Biographical and Administrative History. We haven’t yet explored this thoroughly, but with URIs for archival creators and URIs for administrative and biographical histories, one’s thoughts start to turn to name authorities and EAC-CPF (Encoded Archival Context – Corporate Bodies, Persons and Families – a means to markup information about archival creators in XML). We are not looking at creating EAC descriptions, but it would be good to keep in line with this in whatever ways we can in order to facilitate the subsequent creation of EAC records, or incorporation of our data into EAC records.

We will soon be able to share our current data model, so keep an eye on our blog. We welcome any feedback that the community might have.