dialled.ca
Inspired by Collaborative Writing Resource Centre’s map of Canadian libraries, archives, and museums: 4,798 institutions with locations and URLs
Meshed perfectly with my drive to connect library resources with Canadians by exposing structured data in the web pages describing those resources
Every resource (expressed as a bibliographic description) linked to their instances Every instance linked to the offering library Every library, its URL!
Whereas I was working from the bottom up, the problem I eventually ran into was a lack of description of libraries in a linked open data form
There is the OCLC WorldCat Registry - but it is uneven (only two branches of Toronto Public Library), frequently out of date or inaccurate, and has a non-commercial license
There is the Library and Archives Canada directory - but it holds only one entry for Toronto Public Library (well, there are 8 or so, but they are all superceded by the single TPL entry)
So CWRC’s work was fantastic to see — but it was a snapshot in time dating back to mid-June. And we know everything suffers linkrot on the web.
And it was not granular; while the CWRC data provided lat/long coordinates, those were based entirely at the city level, as were the addresses. If you want to see what branch is closest to you on the CWRC map, you can’t.
But the CWRC data does include URLs… precious, precious URLs. Meaning that we could crawl the URLs and find out what they told us.
And CWRC very generously made the JSON files that include all of the data that they pulled together available; they’re linked right from their web page, one file per province.
So I starting writing some Python scripts. One just downloads and assembles the CWRC JSON files. The current main script runs through the JSON, finds each URL, and tries to download the HTML at the other end. Then it tries to parse any structured data out of the HTML. That script leans heavily on Python libraries like Requests, html5lib, and RDFLib, so it’s only a few hundred lines long.
That linkrot? Only 3,141 pages could be retrieved, and of those, 661 of the URLs redirected. And many of the branches were represented by a single URL for the entire system.
191 pages have microdata (itemscope / itemtype / itemprop) 197 pages have RDFa (typeof / property) 112 pages contain JSON-LD (<script type="application/ld+json>)
287 pages include schema.org
3 pages — one of which is an archives — use the schema:Library type. None use schema:Museum or the emerging schema:Archives. 47 use schema:LocalBusiness 15 use schema:Organization 44 use schema:Event
Current status: Very rough.
Needs URLs (for example, disambiguated 13 PEI libs through a scraping script & manual updates) — and some redirects now point to the overall organization instead of the library-specific page. Ah web redesigns… what do you mean, persistent URIs
Needs libraries to embed useful RDFa, JSON-LD, or microdata in their pages
Needs to pull the address information, geocode it, and integrate more accurate lat/long coordinates.
Prior work: Adrian Pohl at hbz — Hochschulbibliothekszentrum des Landes NRW launched http://beta.lobid.org/organisations way back in 2010; one of the benefits of a centralized ILL service is that he had access to relatively complete data for each of the participating libraries. But he and his team have put together an impressive set of linked data.
So what can you do? You can correct or improve the URLs that are listed. If you work for a consortium—say, something like the Southern Ontario Library Service, or CARL, or the Ontario Library Association—maybe you could point at whatever page you undoubtedly already have that lists all of the libraries and provides their up-to-date URLs; that would be a great help.
You can add linked data to your library homepage. I have some templates that might help you with RDFa: a static HTML one, and a Drupal template that offers user-friendly widgets for configuring opening hours and contact information.
And I wrote an article that leads you through the process of enriching your library homepage with linked data, including location, hours, and contact information, all the way through marking up events.
This can include each page in your catalogue, if you’ve gone to the trouble of embedding linked data in those (or your institutional repository, or digital archives or exhibits). Oh hey, I’ve written up self-guided tutorials on how to mark those up, too! And (sneak peek), I’m going to be leading a hands-on workshop on May 12 at LODLAM TO. http://lodlamto.ca y’all.
And if you really want to go down the rabbithole, you can integrate a sitemap—decades-old technology for describing the relevant pages of your website.
And once you’ve created the sitemap, you can advertise it in the robots.txt file for your site. sitemaps and robots.txt aren’t just for commercial search engines! Sitemaps and robots.txt files are described in my article too.
Maybe you want to play with the data that I’ve collected so far? Sure thing, I’ve made it available as a single Turtle file, ready to be loaded into your favourite triple store and SPARQL-queried into giving you the answers you’re looking for.
Still lots of noise: statements like "<foo> a foaf:Image" and "<foo> xhtml:icon </bar>" are not particularly useful.
Need to follow sameAs declarations to pull in more relevant data; for example, the Laurentian home page does not directly include operating hours or branch relationships, but uses sameAs to point to the Evergreen-generated library page that does include and embed this information.
Having a snazzy web front end, instead of a static HTML site with a TTL dump, would be nice too.