Google announced yesterday that it is the end of the line for Freebase, and they have “decided to help transfer the data in Freebase to Wikidata, and in mid-2015 we’ll wind down the Freebase service as a standalone project”.
As well as retiring access for data creation and reading, they are also retiring API access – not good news for those who have built services on top of them. The timetable they shared for the move is as follows:
Before the end of March 2015
– We’ll launch a Wikidata import review tool
– We’ll announce a transition plan for the Freebase Search API & Suggest Widget to a Knowledge Graph-based solution
March 31, 2015
– Freebase as a service will become read-only
– The website will no longer accept edits
– We’ll retire the MQL write API
June 30, 2015
– We’ll retire the Freebase website and APIs
– The last Freebase data dump will remain available, but developers should check out the Wikidata dump
The crystal ball gazers could probably have predicted a move such as this when Google employed, the then lead of Wikidata, Denny Vrandečić a couple of years back. However they could have predicted a load of other outcomes too. 😉
In the long term this should be good news for Wikidata, but in the short term they may have a severe case of indigestion as they attempt to consume data that will, in some estimations, treble the size of Wikidata adding about 40 million Freebase facts into its current 12 million. It won’t be a simple copy job.
Loading Freebase into Wikidata as-is wouldn’t meet the Wikidata community’s guidelines for citation and sourcing of facts — while a significant portion of the facts in Freebase came from Wikipedia itself, those facts were attributed to Wikipedia and not the actual original non-Wikipedia sources. So we’ll be launching a tool for Wikidata community members to match Freebase assertions to potential citations from either Google Search or our Knowledge Vault, so these individual facts can then be properly loaded to Wikidata.
There are obvious murmurings on the community groups about things such as how strict the differing policies for confirming facts are, and how useful the APIs are. There are bound to be some hiccups on this path – more of an arranged marriage than one of love at first sight between the parties.
I have spent many a presentation telling the world how Google have based their Knowledge Graph on the data from Freebase, which they got when acquiring Metaweb in 2010.
So what does this mean for the Knowledge Graph? I believe it is a symptom of the Knowledge Graph coming of age as a core feature of the Google infrastructure. They have used Freebase to seed the Knowledge Graph, but now that seed has grow into a young tree fed by the twin sources of Google search logs, and the rich nutrients delivered by Schema.org structured data embedded in millions of pages on the web. Following the analogy, the seed of Freebase, as a standalone project/brand, just doesn’t fit anymore with the core tree of knowledge that Google is creating and building. No coincidence that they’ll “announce a transition plan for the Freebase Search API & Suggest Widget to a Knowledge Graph-based solution”.
As for Wikidata, if the marriage of data is successful, it will establish it as the source for open structured data on the web and for facts within Wikipedia.
As the live source for information that will often be broader than the Wikipedia it sprang from, I suspect Wikidata’s rise will spur the eventual demise of that other source of structured data from Wikipedia – DBpedia. How in the long term will it be able to compete, as a transformation of occasional dumps of Wikipedia, with a live evolving broader source? Such a demise would be a slow process however – DBpedia has been the de facto link source for such a long time, its URIs are everywhere!
However you see the eventual outcomes for Frebase, Wikidata, and DBpedia, this is big news for structured data on the web.
It is one thing to have a vision, regular readers of this blog will know I have them all the time, its yet another to see it starting to form through the mist into a reality. Several times in the recent past I have spoken of the some of the building blocks for bibliographic data to play a prominent part in the Web of Data. The Web of Data that is starting to take shape and drive benefits for everyone. Benefits that for many are hiding in plain site on the results pages of search engines. In those informational panels with links to people’s parents, universities, and movies, or maps showing the location of mountains, and retail outlets; incongruously named Knowledge Graphs.
OK, you may say, we’ve heard all that before, so what is new now?
As always it is a couple of seemingly unconnected events that throw things into focus.
Event 1: An article by David Weinberger in the DigitalShift section of Library Journal entitled Let The Future Go. An excellent article telling libraries that they should not be so parochially focused in their own domain whilst looking to how they are going serve their users’ needs in the future. Get our data out there, everywhere, so it can find its way to those users, wherever they are. Making it accessible to all. David references three main ways to provide this access:
APIs – to allow systems to directly access our library system data and functionality
Linked Data – can help us open up the future of libraries. By making clouds of linked data available, people can pull together data from across domains
The Library Graph – an ambitious project libraries could choose to undertake as a group that would jump-start the web presence of what libraries know: a library graph. A graph, such as Facebook’s Social Graph and Google’s Knowledge Graph, associates entities (“nodes”) with other entities
(I am fortunate to be a part of an organisation, OCLC, making significant progress on making all three of these a reality – the first one is already baked into the core of OCLC products and services)
It is the 3rd of those, however, that triggered recognition for me. Personally, I believe that we should not be focusing on a specific ‘Library Graph’ but more on the ‘Library Corner of a Giant Global Graph’ – if graphs can have corners that is. Libraries have rich specialised resources and have specific needs and processes that may need special attention to enable opening up of our data. However, when opened up in context of a graph, it should be part of the same graph that we all navigate in search of information whoever and wherever we are.
ZBW contributes to WorldCat, and has 1.2 million oclc numbers attached to it’s bibliographic records. So it seemed interesting, how many of these editions link to works and furthermore to other editions of the very same work.
The post is interesting from a couple of points of view. Firstly the simple steps they took to get at the data, really well demonstrated by the command-line calls used to access the data – get OCLCNum data from WorldCat.or in JSON format – extract the schema:exampleOfWork link to the Work – get the Work data from WorldCat, also in JSON – parse out the links to other editions of the work and compare with their own data. Command-line calls that were no doubt embedded in simple scripts.
Secondly, was the implicit way that the corpus of WorldCat Work entity descriptions, and their canonical identifying URIs, is used as an authoritative hub for Works and their editions. A concept that is not new in the library world, we have been doing this sort of things with names and person identities via other authoritative hubs, such as VIAF, for ages. What is new here is that it is a hub for Works and their relationships, and the bidirectional nature of those relationships – work to edition, edition to work – in the beginnings of a library graph linked to other hubs for subjects, people, etc.
The ZBW Labs experiment is interesting in its own way – simple approach enlightening results. What is more interesting for me, is it demonstrates a baby step towards the way the Library corner of that Global Web of Data will not only naturally form (as we expose and share data in this way – linked entity descriptions), but naturally fit in to future library workflows with all sorts of consequential benefits.
The experiment is exactly the type of initiative that we hoped to stimulate by releasing the Works data. Using it for things we never envisaged, delivering unexpected value to our community. I can’t wait to hear about other initiatives like this that we can all learn from.
So who is going to be doing this kind of thing – describing entities and sharing them to establish these hubs (nodes) that will form the graph. Some are already there, in the traditional authority file hubs: The Library of Congress LC Linked Data Service for authorities and vocabularies (id.loc.gov), VIAF, ISNI, FAST, Getty vocabularies, etc.
As previously mentioned Work is only the first of several entity descriptions that are being developed in OCLC for exposure and sharing. When others, such as Person, Place, etc., emerge we will have a foundation of part of a library graph – a graph that can and will be used, and added to, across the library domain and then on into the rest of the Global Web of Data. An important authoritative corner, of a corner, of the Giant Global Graph.
As I said at the start these are baby steps towards a vision that is forming out of the mist. I hope you and others can see it too.
Regular readers of this blog may well know I am an enthusiast for Schema.org – the generic vocabulary for describing things on the web as structured data, backed by the major search engines Google, Bing, Yahoo! & Yandex. When I first got my head around it back in 2011 I soon realised it’s potential for making bibliographic resources, especially those within libraries, a heck of a lot more discoverable. To be frank library resources did not, and still don’t, exactly leap in to view when searching the web – a bit of a problem when most people start searching for things with Google et al – and do not look elsewhere.
Schema.org as a generic vocabulary to describe most stuff, easily embedded in your web pages, has been a great success. As was reported by Google’s R.V. Guha, at the recent Semantic Technology and Business Conference in San Jose, a sample of 12B pages showed approximately 21% containing Schema.org markup. Right from the beginning, however, I had concerns about its applicability to the bibliographic world – great start with the Book type, but there were gaps the coverage for such things as journal issues & volumes, multi-volume works, citations, and the relationship between a work and its editions. Discovering others shared my combination of enthusiasm and concerns, I formed a W3C Community Group – Schema Bib Extend – to propose some bibliographic focused extensions to Schema.org. Which brings me to the events behind this post…
The SchemaBibEx group have had several proposals accepted over the last couple of years, such as making the [commercial] Offer more appropriate for describing loanable materials, and broadening of the citation property. Several other significant proposals were brought together in a package which I take great pleasure in reporting was included in the latest v1.9 release of Schema.org. For many in our group these latest proposals were a long time coming after their initial proposal. Although frustrating, the delays were symptomatic of a very healthy process.
Although the number of new types and properties are small, their addition to Schema opens up potential for much better description of periodicals and creative work relationships. To introduce the background to this, SchemaBibEx member Dan Scott and I were invited to jointly post on the Schema.org Blog.
So, another step forward for Schema.org. I believe that is more than just a step however, for those wishing to make the bibliographic resources more visible on the Web. There as been some criticism that Schema.org has been too simplistic to be able represent some of the relationships and subtleties from our world. Criticism that was not unfounded. Now with these enhancements, much of these criticisms are answered. There is more to do, but the major objective of the group that proposed them has been achieved – to lay the broad foundation for the description of bibliographic, and creative work, resources in sufficient detail for them to be understood by the search engines to become part of their knowledge graphs. Of course that is not the final end we are seeking. The reason we share data is so that folks are guided to our resources – by sharing, using the well understood vocabulary, Schema.org.
Examples of a conceptual creative work being related to its editions, using exampleOfWork and workExample, have been available for some time. In anticipation of their appearance in Schema, they were introduced into the OCLC WorldCat release of 194 million Work descriptions (for example: http://worldcat.org/entity/work/id/1363251773) with the inverse relationship being asserted in an updated version of the basic WorldCat linked data that has been available since 2012.
A couple of months back I spoke about the preview release of Works data from WorldCat.org. Today OCLC published a press release announcing the official release of 197 million descriptions of bibliographic Works.
A Work is a high-level description of a resource, containing information such as author, name, descriptions, subjects etc., common to all editions of the work. The description format is based upon some of the properties defined by the CreativeWork type from the Schema.org vocabulary. In the case of a WorldCat Work description, it also contains [Linked Data] links to individual, OCLC numbered, editions already shared from WorldCat.org.
These links (URIs) lead, where available, to authoritative sources for people, subjects, etc. When not available, placeholder URIs have been created to capture information not yet available or identified in such authoritative hubs. As you would expect from a linked data hub the works are available in common RDF serializations – Turtle, RDF/XML, N-Triples, JSON-LD – using the Schema.org vocabulary – under an open data license.
The obvious question is “how do I get a work id for the items in my catalogue?”. The simplest way is to use the already released linked data from WorldCat.org. If you have an OCLC Number (eg. 817185721) you can create the URI for that particular manifestation by prefixing it with ‘http://worldcat.org/oclc/’ thus: http://worldcat.org/oclc/817185721
In the linked data that is returned, either on screen in the Linked Data section, or in the RDF in your desired serialization, you will find the following triple which provides the URI of the work for this manifestation:
To quote Neil Wilson, Head of Metadata Services at the British Library:
With this release of WorldCat Works, OCLC is creating a significant, practical contribution to the wider community discussion on how to migrate from traditional institutional library catalogues to popular web resources and services using linked library data. This release provides the information community with a valuable opportunity to assess how the benefits of a works-based approach could impact a new generation of library services.
This is a major first step in a journey to provide linked data views of the entities within WorldCat. Looking forward to other WorldCat entities such as people, places, and events. Apart from major release of linked data, this capability is the result of applying [Big] Data mining and analysis techniques that have been the focus of research and development for several years. These efforts are demonstrating that there is much more to library linked data than the mechanical, record at a time, conversion of Marc records into an RDF representation.
You may find it helpful, in understanding the potential exposed by the release of Works, to review some of the questions and answers that were raised after the preview release.
Personally I am really looking forward to hearing about the uses that are made of this data.
One of the most challenging challenges in my evangelism of the benefits of using Schema.org for sharing data about resources via the web is that it is difficult to ‘show’ what is going on.
The scenario goes something like this…..
“Using the Schema.org vocabulary, you embed data about your resources in the HTML that makes up the page using either microdata or RDFa….”
At about this time you usually display a slide showing html code with embedded RDFa. It may look pretty but the chances of more than a few of the audience being able to pick out the schema:Book or sameAs or rdf:type elements out of the plethora of angle brackets and quotes swimming before their eyes is fairly remote.
Having asked them to take a leap of faith that the gobbledegook you have just presented them with, is not only simple to produce but also invisible to users viewing their pages – “but not to Google, which harvest that meaningful structured data from within your pages” – you ask them to take another leap [of faith].
You ask them to take on trust that Google is actually understanding, indexing and using that structured data. At this point you start searching for suitable screen shots of Google Knowledge Graph to sit behind you whilst you hypothesise about the latest incarnation of their all-powerful search algorithm, and how they imply that they use the Schema.org data to drive so-called Semantic Search.
I enjoy a challenge, but I also like to find a better way sometimes. w3
When OCLC first released Linked Data in WorldCat they very helpfully addressed the first of these issues by adding a visual display of the Linked Data to the bottom of each page. This made my job far easier!
But it has a couple of downsides. Firstly it is not the prettiest of displays and is only really of use to those interested in ‘seeing’ Linked Data. Secondly, I believe it creates an impression to some that, if you want Google to grab structured data about resources, you need to display a chunk of gobbledegook on your pages.
That simple way to easily show someone the data embedded in a page, is a great aid to understanding for those new to the concept. But that is not all. This excellent little extension has a couple of extra tricks up its sleeve.
It includes a visualisation of the [Linked Data] graph of relationships – the structure of the data. Clicking on any of the nodes of the display, causes the value of the subject, predicate, or object it represents to be displayed below the image and the relevant row(s) in the list of triples to be highlighted. As well as all this, there is a ‘Show Turtle’ button, which does just as you would expect opening up a window in which it has translated the triples into Turtle – Turtle being (after a bit of practise) the more human friendly way of viewing or creating RDF.
Green Turtle is a useful little tool which I would recommend to visualise microdata and RDFa, be it using the Schema.org vocabulary or not. I am already using it on WorldCat in preference to scrolling to the bottom of the page to click the Linked Data tab.
Custom Searches that know about Schema! Google have recently enhanced the functionality of their Custom Search Engine (CSE) to enable searching by Schema.org Types. Try out this example CSE which only returns results from WorldCat.org which have been described in their structured data as being of type schema:Book.
A simple yet powerful demonstration that not only are Google harvesting the Schema.org Linked Data from WorldCat, but they are also understanding it and are visibly using it to drive functionality.
Instead of keeping the answers within individual email threads, I thought they may be of interest to a wider audience:
QI don’t see anything that describes the criteria for “workness.” “Workness” definition is more the result of several interdependent algorithmic decision processes than a simple set of criteria. To a certain extent publishing the results as linked data was the easy (huh!) bit. The efforts to produce these definitions and their relationships are the ongoing results of a research process, by OCLC Research, that has been in motion for several years, to investigate and benefit from FRBR. You can find more detail behind this research here: http://www.oclc.org/research/activities/frbr.html?urlm=159763
Q Defining what a “work” is has proven next to impossible in the commercial world, how will this be more successful? Very true for often commercial and/or political, reasons previous initiatives in this direction have not been very successful. OCLC make no broader claim to the definition of a WorldCat Work, other than it is the result of applying the results of the FRBR and associated algorithms, developed by OCLC Research, to the vast collection of bibliographic data contributed, maintained, and shared by the OCLC member libraries and partners.
QWill there be links to individual ISBN/ISNI records?
ISBN – ISBNs are attributes of manifestation [in FRBR terms] entities, and as such can be found in the already released WorldCat Linked Data. As each work is linked to its related manifestation entities [by schema:workExample] they are therefore already linked to ISBNs.
ISNI – ISNI is an identifier for a person and as such an ISNI URI is a candidate for use in linking Works to other entity types. VIAF URIs being another for Person/Organisation entities which, as we have the data, we will be using. No final decisions have been made as to which URIs we use and as to using multiple URIs for the same relationship. Do we Use ISNI, VIAF, & Dbpedia URIs for the same person, or just use one and rely on interconnection between the authoritative hubs, is a question still to be concluded.
Q Can you say more about how the stable identifiers will be managed as the grouping of records that create a work change? You correctly identify the issue of maintaining identifiers as work groups split & merge. This is one of the tasks the development team are currently working on as they move towards full release of this data over the coming weeks. As I indicated in my blog post, there is a significant data refresh due and from that point onwards any changes will be handled correctly.
Q Is there a bulk download available? No there is no bulk download available. This is a deliberate decision for several reasons.
Firstly this is Linked Data – its main benefits accrue from its canonical persistent identifiers and the relationships it maintains between other identified entities within a stable, yet changing, web of data. WorldCat.org is a live data set actively maintained and updated by the thousands of member libraries, data partners, and OCLC staff and processes. I would discourage reliance on local storage of this data, as it will rapidly evolve and become out of synchronisation with the source. The whole point and value of persistent identifiers, which you would reference locally, is that they will always dereference to the current version of the data.
Q Where should bugs be reported? Today, you can either use the comment link from the Linked Data Explorer or report them to firstname.lastname@example.org. We will be building on this as we move towards full release.
QThere appears to be something funky with the way non-existent IDs are handled. You have spotted a defect! – The result of access to a non established URI should be no triples returned with that URI as subject. How this is represented will differ between serialisations. Also you would expect to receive a http status of 404 returned.
QIt’s wonderful to see that the data is being licensed ODC-BY, but maybe assertions to that effect should be there in the data as well?. The next release of data will be linked to a void document providing information, including licensing, for the dataset.
Q How might WorldCat Works intersect with the BIBFRAME model? – these work descriptions could be very useful as a bf:hasAuthority for a bf:Work. The OCLC team monitor, participate in, and take account of many discussions – BIBFRAME, Schema.org, SchemaBibEx, WikiData, etc. – where there are some obvious synergies in objectives, and differences in approach and/or levels of detail for different audiences. The potential for interconnection of datasets using sameAs, and other authoritative relationships such as you describe is significant. As the WorldCat data matures and other datasets are published, one would expect initiatives from many in starting to interlink bibliographic resources from many sources.
Q Will your team be making use of ISTC? Again it is still early for decisions in this area. However we would not expect to store the ISTC code as a property of Work. ISTC is one of many work based data sets, from national libraries and others, that it would be interesting to investigate processes for identifying sameAs relationships between.
The answer to the above question stimulated a follow-on question based upon the fact that ISTC Codes are allocated on a language basis. In FRBR terms language of publication is associated with the Expression, not the Work level description. As such therefore you would not expect to find ISTC on a ‘Work’ – My response to this was:
Note that the Works published from WorldCat.org are defined as instances of schema:CreativeWork.
What you say may well be correct for FRBR, but the the WorldCat data may not adhere strictly to the FRBR rules and levels. I say ‘may not’ as we are still working the modelling behind this and a language specific Work may become just an example of a more general Work – there again it may become more Expression-like. There is a balance to be struck between FRBR rules and a wider, non-library, understanding.
Q Which triplestore are you using? We are not using a triplestore. Already, in this early stage of the journey to publish linked data about the resources within WorldCat, the descriptions of hundreds of millions of entities have been published. There is obvious potential for this to grow to many billions. The initial objective is to reliably publish this data in ways that it is easily consumed, linked to, and available in the de facto linked data serialisations. To achieve this we have put in place a simple very scalable, flexible infrastructure currently based upon Apache Tomcat serving up individual RDF descriptions stored in Apache HBase (built on top of Apache Hadoop HDFS). No doubt future use cases will emerge, which will build upon this basic yet very valuable publishing of data, that will require additional tools, techniques, and technologies to become part of that infrastructure over time. I know the development team are looking forward to the challenges that the quantity, variety, and always changing nature of data within WorldCat will provide for some of the traditional [for smaller data sets] answers to such needs.
As an aside, you may be interested to know that significant use is made of the map/reduce capabilities of Apache Hadoop in the processing of data extracted from bibliographic records, the identification of entities within that data, and the creation of the RDF descriptions. I think it is safe to say that the creation and publication of this data would not have been feasible without Hadoop being part of the OCLC architecture.
Hopefully this background will help those interested in the process. When we move from preview to a fuller release I expect to see associated documentation and background information appear.
I have just been sharing a platform, at the OCLC EMEA Regional Council Meeting in Cape Town South Africa, with my colleague Ted Fons. A great setting for a great couple of days of the OCLC EMEA membership and others sharing thoughts, practices, collaborative ideas and innovations.
Ted and I presented our continuing insight into The Power of Shared Data, and the evolving data strategy for the bibliographic data behind WorldCat. If you want to see a previous view of these themes you can check out some recordings we made late last year on YouTube, from Ted – The Power of Shared Data – and me – What the Web Wants.
Today, demonstrating on-going progress towards implementing the strategy, I had the pleasure to preview two upcoming significant announcements on the WorldCat data front:
The release of 194 Million Open Linked Data Bibliographic Work descriptions
The WorldCat Linked Data Explorer interface
A Work is a high-level description of a resource, containing information such as author, name, descriptions, subjects etc., common to all editions of the work. The description format is based upon some of the properties defined by the CreativeWork type from the Schema.org vocabulary. In the case of a WorldCat Work description, it also contains [Linked Data] links to individual, oclc numbered, editions already shared in WorldCat. Let’s take a look at one – try this: http://worldcat.org/entity/work/id/12477503
You will see, displayed in the new WorldCat Linked Data Explorer, a html view of the data describing ‘Zen and the art of motorcycle maintenance’. Click on the ‘Open All’ button to view everything. Anyone used to viewing bibliographic data will see that this is a very different view of things. It is mostly URIs, the only visible strings being the name or description elements. This is not designed as an end-user interface, it is designed as a data exploration tool. This is highlighted by the links at the top to alternative RDF serialisations of the data – Turtle, N-Triple, JSON-LD, RDF/XML.
Why is this a preview? Can I usefully use the data now? Are a couple of obvious questions for you to ask at this time.
This is the first production release of WorldCat infrastructure delivering linked data. The first step in what will be an evolutionary, and revolutionary journey, to provide interconnected linked data views of the rich entities (works, people, organisations, concepts, places, events) captured in the vast shared collection of bibliographic records that makes up WorldCat. Mining those, 311+ million, records is not a simple task, even to just identify works. It takes time, and a significant amount of [Big Data] computing resources. One of the key steps in this process is to identify where they exist connections between works and authoritative data hubs, such as VIAF, FAST, LCSH, etc. In this preview release, it is some of those connections that are not yet in place.
What you see in their place at the moment is a link to, what can be described as, a local authority. These are exemplified by what the data geeks call a hash-URI as its identifier. http://experiment.worldcat.org/entity/work/data/12477503#Person/pirsig_robert for example is such an identifier, constructed from the work URI and the person name. Over the next few weeks, where the information is available, you would expect to see this link replaced by a connection to VIAF, such as this: http://viaf.org/viaf/78757182.
So, can I use the data? – Yes, the data is live, and most importantly the work URIs are persistent. It is also available under an open data license (ODC-BY).
In a very few weeks, once the next update to the WorldCat linked data has been processed, you will find that links to works will be embedded in the already published linked data. For example you will find the following in the data for OCLC number 53474380:
What is next on the agenda? As described, within a few weeks, we expect to enhance the linking within the descriptions and provide links from the oclc numbered manifestations. From then on, both WorldCat and others will start to use WorldCat Work URIs, and their descriptions, as a core stable foundations to build out a web of relationships between entities in the library domain. It is that web of data that will stimulate the sharing of data and innovation in the design of applications and interfaces consuming the data over coming months and years.
As I said on the program today, we are looking for feedback on these releases.
We as a community are embarking on a new journey with shared, linked data at its heart. Its success will be based upon how that data is exposed, used, and the intrinsic quality of that data. Experience shows that a new view of data often exposes previously unseen issues, it is just that sort of feedback we are looking for. So any feedback on any aspect of this will be more than welcome.
I am excitedly looking forward to being able to comment further as this journey progresses.
The Art & Architecture Thesaurus is a reference of over 250,000 terms on art and architectural history, styles, and techniques. I’m sure this will become an indispensible authoritative hub of terms in the Web of Data to assist those describing their resources and placing them in context in that Web.
This is the fist step in an 18 month process to release four vocabularies – the others being The Getty Thesaurus of Geographic Names (TGN)®, The Union List of Artist Names®, and The Cultural Objects Name Authority (CONA)®.
A great step from Getty. I look forward to the others appearing over the months and seeing how rapidly their use is made across the web.