Initially this entry only containd the SPARQL endpoints and not downloads of the full datasets. After some enquiries from Naomi Lillie the entry was updated with links to the some more information and bulk downloads at: http://www.bne.es/es/Catalogos/DatosEnlazados/DescargaFicheros/

This library dataset is particularly interesting as it is not a ‘straightforward’ dump of bibliographic records. This is best explained by Karen Coyle in her blogpost.

For a BibServer import,  the implications are that we have to distinguish the types of record that is read by the importing script and take the relevant action before building the BibJSON entry. Fortunately the datadump was made as N-Triples already, so we did not have to pre-process the large datafile (4.9GB) in the same manner as we did with the German National Library dataset.

The Python script to perform the reading of the datafile can be viewed at https://gist.github.com/3225004

A complicating matter from a data wrangler’s point of view is that the field names are based on IFLA Standards, which are numeric codes and not ‘guessable’ English terms like DublinCore fields for example. This is more correct from an international and data quality point of view, but does make the initial mapping more time consuming.

 So when mapping a data item like https://gist.github.com/3225004#file_sample.nt we need to dereference each fieldname and map it to the relevant BibJSON entry.

As we identify more Linked Open Data National Bibliographies, these experiments will be continued under the http://nb.bibsoup.net/ BibServer instance.

The Discussion List has been busy too:

• Further to David Weinbergers’s pointer that Harvard released 12 million bibliographic records with a CC0 licence, Rufus Pollock created a collection on the DataHub and added it to the Biblio section for easy of reference

• Rufus also noticed that OCLC had issued their major release of VIAF, meaning that millions of author records are now available as Open Data (under Open Data Commons Attribution license), and updated the DataHub dataset to reflect this

• Peter Murray-Rust noted that Nature has made its metadata Open CC0

• David Shotton promoted the International Workshop on Contributorship and Scholarly Attribution at Harvard, and prepared a handy guide for attribution of submissions

• Adrian Pohl circulated a call for participation for the SWIB12 “Semantic Web in Bibliotheken” (Semantic Web in Libraries) Conference in Cologne, 26-28 November this year, and hosted the monthly Working Group call

• Lars Aronsson looked at multivolume works, asking whether the OpenLibrary can create and connect records for each volume. HathiTrust and Gallica were suggested as potential tools in collating volumes, and the barcode (containing information populated by the source library) was noted as being invaluable in processing these

• Sam Leon explained that TEXTUS would be integrating BibSever facet view and encouraged people to have a look at the work so far; Tom Oinn highlighted the collaboration between Enriched BibJSON and TEXTUS, and explained that he would be adding a ‘TEXTUS’ field to BibJSON for this purpose

• Sam also circulated two tools for people to test, Pundit and Korbo, which have been developed out of Digitised Manuscripts to Europeana (DM2E)

• Jenny Molloy promoted the Open Science Hackday which took place last week – see below for a snap-shot courtesy of @OKFN:

In related news, Peter Murray-Rust is continuing to advocate the cause of open data – do have a read of the latest posts on his blog to see how he’s getting on.

The Open Biblio community continues to be invaluable to the Open GLAM, Heritage, Access and other groups too and I would encourage those interested in such discussions to join up at the OKFN Lists page.

We are happy that the German National Library recently released the German National Bibliography as Linked Open Data, see (announcement). At the #bibliohack this week we worked on getting the data into a BibServer instance. Here, we want to share our experiences in trying to re-use this dataset.

Parsing large turtle files: problem and solution

The raw data file is 1.1GB in a compressed format – unzipped it is a 6.8 GB turtle file.
Working with this file is unwieldy, it can not be read into memory or converted with tools like rapper (which only works for turtle files up to 2 GB, see this mail thread). Thus, it would be nice if the German National Library could either provide one big N-Triples file that is better for streaming processing or provide a number of smaller turtle files.

Our solution to get the file into a workable form is to make a small Python script that is Turtle syntax aware, to split the file into smaller pieces. You can’t use the standard UNIX split command, as each snippet of the split file also needs the prefix information at the top and we do not want to split an entry in the middle, losing triples.

See a sample converted N-Triples file from a turtle snippet.

Converting the N-Triples to BibJSON

After this, we started working on parsing an example N-Triples file to convert the data to BibJSON. We haven’t gotten that far, though. See https://gist.github.com/2928984#file_ntriple2bibjson.py for the resulting code (work in progress).

Problems

We noted problems with some properties that we like to document here as feedback for the German National Library.

Heterogeneous use of dcterms:extent

The dcterms:extent property is used in many different ways, thus we are considering to omit it in the conversion to BibJSON. Some example values of this property: “Mikrofiches”, “21 cm”, “CD-ROMs”, “Videokassetten”, “XVII, 330 S.”. Probably it would be the more appropriate choice to use dcterms:format for most of these and to limit the use of dcterms:extent to pagination information and duration.

URIs that don’t resolve

We stumbled over some URIs that don’t resolve, whether you order RDF or HTML in the accept header. Examples: http://d-nb.info/019673442, http://d-nb.info/019675585, http://d-nb.info/011077166

Also, DDC URIs that are connected to a resource with dcters:subject don’t resolve, e.g. http://d-nb.info/ddc-sg/070.

Footnote

At a previous BibServer hackday, we loaded the Britsh National Bibliography data into BibServer. This was a similar problem, but as the data was in RDF/XML we could directly use the built-in Python XML streaming parser to convert the RDF data into BibJSON.
See: https://gist.github.com/1731588 for the source.

BiblioHack has drawn to a close and the results of two days’ hard labour are in:

A Bibliographic Toolkit

Utilising BibServer

Peter Murray-Rust reported back on what was planned, what was done, and the overlap between the two! The priority was cleaning up the process for setting up BibServers and getting them running on different architectures. (PubCrawler was going to be run on BibServer but currently it’s not working). Yesterday’s big news was that Nature has released 30 million references or thereabouts – this furthers the cause of scholarly literature whereby we, in principle, can index records rather than just corporate organisations being able / permitted to do so. National Bibliographies have been put on BibSoup – UK (‘BL’), Germany, Spain and Sweden – with the technical problem character encodings raising its head (UTF8 solves this where used). Also, BibSoup is useful for TEXTUS so the overall ‘toolkit’ approach is reinforced!

Open Access Index

Emanuil Tolev presented on ACat – Academic Catalogue. The first part of an index is having things to access – so gathering about 55,000 journals was a good start! Using Elastic Search within these journals will give list of contents which will then provide lists of articles (via facet view), then other services will determine licensing / open access information (URL checks assisted in this process). The ongoing plan is to use this tool to ascertain licensing information for every single record in the world. (Link to ACat to follow).

Annotation Tools

Tom Oinn talked about the ideas that have come out of discussions and hacking around annotators and TEXTUS. Reading lists and citation management is a key part of what TEXTUS is intended to assist with, so the plan is for any annotation to be allowed to carry a citation – whether personal opinion or related record. Personalised lists will come out of this and TEXTUS should become a reference management tool in its own right. Keep your eye on TEXTUS for the practical applications of these ideas!

Note: more detailed write-ups will appear courtesy of others, do watch the OKFN blog for this and all things open…

Postscript: OKFN blog post here

Huge thanks to all those who participated in the event – your ideas and enthusiasm have made this so much fun to be involved with.

Also thanks to those who helped run the event, visible or behind-the-scenes, particularly Sam Leon.

Here’s to the next one

• Peter Murray-Rust revisited the overarching theme of ‘A Bibliographic Toolkit’ and the BibServer sub-group’s specific work on adding datasets and easily deploying BibServer; Adrian Pohl followed up to explain that he would be developing a National Libraries BibServer.
• Tom Oinn explained the Annotation Tools sub-groups’s work on developing annotation tools – ie TEXTUS – looking at adding fragments of text, with your own comments and metadata linked to it, which then forms BibSoup collections. Collating personalised references is enhanced with existing search functionality, and reading lists with annotations can refer to other texts within TEXTUS.
• Mark MacGillivray presented the 3rd group’s work on an Open Access Index. This began with listing all the journals that can be found in the whole world, with the aim of identifying the licence of each article. They have been scraping collections (eg PubMed) and gathering journals – at the time of speaking they had around 50,000+! The aim is to enable a crowd-sourced list of every journal in the world which, using PubCrawler, should provide every single article in the world.

With just 5 hours left before stopping to gather thoughts, write-up and feedback to the rest of the group, it will be very interesting to see the result…

The event attendees started the day all together, both hackers and workshop / seminar attendees, and Sam introduced the purpose of the day as follows: coders – to build tools and share ideas about things that will make our shared cultural heritage and knowledge commons more accessible and useful; non-coders – to get a crash course in what openness means for galleries, libraries, archives and museums, why it’s important and how you can begin opening up your data; everyone – to get a better idea about what other people working in your domain do and engender a better understanding between librarians, academics, curators, artists and technologists, in order to foster the creation of better, cooler tools that respond to the needs of our communities.

The hackers began the day with an overview of what a hackathon is for and how it can be run, as presented by Mahendra Mahey, and followed with lightning talks as follows:

• Talk 1 Peter Murray Rust & Ross Mounce – Content and Data Mining and a PDF extractor
• Talk 2 Mike Jones – the m-biblio project
• Talk 4 Ian Stuart – ORI/RJB (formerly OA-RJ)
• Talk 5 Etienne Posthumus – Making a BibServer Parser
• Talk 6 Emanuil Tolev – IDFind – identifying identifiers (“Feedback and real user needs won’t gather themselves”)
• Talk 7 Mark MacGillivray – BibServer – what the project has been doing recently, how that ties into the open access index idea.
• Talk 8 Tom Oinn – TEXTUS
• Talk 9 Simone Fonda – Pundit – collaborative semantic annotations of texts (Semantic Web-related tool)
• Talk 10 Ian Stuart – The basics of Linked Data

We decided we wanted to work as a community, using our different skills towards one overarching goal, rather than breaking into smaller groups with separate agendas. We formed the central idea of an ‘open bibliographic tool-kit’ and people identified three main areas to hack around, playing to their skills and interests:

• Utilising BibServer – adding datasets and using PubCrawler
• Creating an Open Access Index
• Developing annotation tools

At this point we all broke for lunch, and the workshoppers and hackers mingled together. As hoped, conversations sprung up between people from the two different groups and it was great to see suggestions arising from shared ideas and applications of one group being explained to the theories of the other.

We re-grouped and the workshop continued until 16.00 – see here for Tim Hodson’s excellent write-up of the event and talks given – when the hackers were joined by some who attended the workshop. Each group gave a quick update on status, to try to persuade the new additions to the group to join their particular work-flow, and each group grew in number. After more hushed discussions and typing, the day finished with a talk from Tara Taubman about her background in the legalities of online security and IP, and we went for dinner. Hacking continued afterwards and we celebrated a hard day’s work down the pub, lookong forward to what was to come.

Day 2 to follow…

Pubcrawler collects bibliographic metadata (author, title, reference, DOI) by indexing journals’ websites in a similar manner to the way in which search engines explore the web to build their indexes. Where possible (which depends on the particular publication) it identifies any supplementary resources associated with a paper, and whether the paper is open access (i.e. readable without a subscription or any other charge) – though it cannot determine the license / conditions of such access.

Pubcrawler was originally developed by Nick Day as part of the CrystalEye project to aggregate published crystallographic structures from the supplementary data to articles on journals’ websites. Since then Pubcrawler has been extended to collect bibliographic metadata and support a wider range of journals than just those containing crystallography. Some of the activities Pubcrawler can currently support are:

• Providing core bibliographic metadata
• Identifying collections of open access articles
• Identifying freely accessible supplementary information, which is often a rich source of scientific data

When pointed at a publisher’s homepage Pubcrawler will generate a list of the journals on the site and then crawl the issues’ tables of contents, recording the bibliographic metadata for the articles that it discovers. Pubcrawler uses a combination of two approaches to crawling a journal: starting at the current issue it can follow links to previous issues, walking the journal’s publication history, and if a journal’s website contains a list of issues it will also use that as a source of pages to crawl. When necessary, such as to identify supplementary resources, Pubcrawler can follow links to individual articles’ splash pages.

Pubcrawler does not index any content that is restricted by a journal’s paywall – it has been designed not to follow such links, and as added protection it is run over a commercial broadband connection, rather than from inside a University network to ensure that it does not receive any kind of privileged access.

While Pubcrawler’s general workflow is the same for any publication, custom parsers are required to extract the metadata and correct links from each website. Generally publishers use common templates for their journals web pages, so a parser only needs to be developed once per publishers, however in some instances, such as where older issues have not been updated to match the current template, a parser may need to support a variety of styles.

Pubcrawler currently has parsers (in varying states of completeness) for a number of publishers (biased by its history of indexing published Crystallographic structures):

• The American Chemical Society (ACS)
• Elsevier
• The International Union of Crystallography (IUCr)
• Nature
• The Royal Society of Chemistry (RSC)
• Springer
• Wiley

And to date it has indexed over 10 million bibliographic records.

There are many other publishers who could be supported by Pubcrawler, they just require parsers to be created for them. Pubcrawler requires two types of maintainance – the general support to keep it running, administer servers etc, that any software requires, and occasional updates to the parsers as journal’s websites change their formatting.

The new scheduled collection uploader is now runnable as a stand-alone tool, to which source URLs can be provided for retrieval, conversion, and upload. Retrieved sources are stored and available from a folder on disk, as are the conversions.

Parsers can now be written in any language and plugged into the ingest functionality – for example, we now have a MARC parser that runs in perl and is usable via ingest.py and available on an instance of BibServer – thanks very much to Ed for that.

In addition, parsers need no longer be ‘parsers’ – we have introduced the concept of scrapers as well. Check out our new Wikipedia parser / scraper, for example; it functions by taking in a search value rather than a URL, then using that to search Wikipedia for relevant references which it downloads, bundles, and converts to a BibJSON collection – this is a really great example that Etienne put together, and it demonstrates a great deal of potential for further parser / scraper development.

See the examples on the BibServer repo for more insight – they are in the parserscrapers_plugins folder, and they are managed by bibserver/ingest.py.

We know documents are now lacking – we have set up an online docs resource but are in the process of writing up to populate it – please check back soon.

As usual, development work is scheduled via the tickets and milestones on our repo. Current efforts are on documentation and adding as many feature requests as possible before our hackathon on June 12th – 14th.

This enables us to keep the core of BibJSON pretty simple whilst also opening up potential for more complex usage where that is required.

Due to this, we no longer use the “namespace” key in BibJSON.

Other changes include usage of “_” prefix on internal keys – so wherever our own database writes info into a record, we prefix it, such as “_id”. Because of this, uploaded BibJSON records can have an “id” key that will work, as well as an “_id” uuid applied by the BibServer system.

For more information, check out BibJSON.org and JSON-LD

Main pages

• Collections visualisation – a smart new graphic on the landing page showing information from new collections

• Improved FAQ section with links to videos (coming soon: links to our new online docs)

Creating collections

• New Wikipedia parser – create a collection based on the references retrievable from Wikipedia for your chosen search value

• Improved collection upload – specify collection information, then view upload tickets to see progress and errors

• ‘Retry’ and other options on particular collection creation attempts are also now available from the tickets page

Search results

• Filter search results by a value range as well as specific values

• Visualise any filter as a bubble chart and select the values you want to search with

• Add / remove available filters and rename filter display names

• Improved layout of record info in search results, including auto-display of the first image referenced in a record – e.g. if there is a link to an image in your record, it is displayed in the search result

Managing and sharing collections

• Collection admin available – save your current display settings as the default for your collection, allow other users to have admin rights on your own collection

• Share any specific searches by providing the URL displayed under the ‘share’ option

• Embed – as the whole front-end of search and collection visualisation is handled by facetview it is possible to embed your collection search in any web page you control; the share / embed option on collection pages provides the code you need to insert to enable this

• Download as BibJSON – a nice new obvious button on each collection provides a link to download your collection as BibJSON

Viewing records

• Improved display of individual records, including search options to discover relevant content online

• EXPERIMENTAL record editing – this has been enabled although still in progress – you can edit the content of a record using a visual display of the keys and values in the record, although functionality for adding new keys does not yet work. However, you can also edit the JSON directly via the options, and try saving that. Be aware – this could damage your records, and of course changes the details from whatever they were in the source content.

Still in development

These ones are not yet available on BibSoup but watch this space:

• Creating new collections on-site – search and find particular records for inclusion in new collections or addition to pre-existing collections. This is not currently possible but we are working on making this an easy process

• Merging collections

• Better user creation and management, plus gravatars

• Additional functionality on record pages – linking out directly to related sources such as PubMed, Total Impact, Service Core etc

We hope you like these changes, and find them useful – do let us know what you think and keep an eye out for the upcoming improvements.