Irish Studies - making library data work harder

Editor's Notes

• #5: Top – James Joyce Library, University College Dublin. I noted that I graduated with a degree in English Language and Literature from UCD. Bottom – the Brazen Head (left) is the oldest pub in Dublin Ireland. This has inspired the Brazen Head (right) in Dublin, Ohio, where OCLC HQ is.
• #6: Finders are experimental work-based systems developed by OCLC Research. This shows works ranked by library holdings.
• #8: The Knowledge Graph that Google began promoting in 2012 is often noted as a model of the power of linked data. Recently, as Knowledge Panels have appeared with more frequency in search engine results, they have been pointed to as exemplars of that power, with libraries, archives and museums aspiring to be reflected in those highlighted views, and with some library systems emulating that type of focus on working in the data.
• #10: Google ‘knows about’ Eva Gore-Booth.
• #11: Google ‘knows about’ Sr Patrick Dun’s Hospital and the Slade School of Fine Art.
• #12: However, Google does not ‘know about’ this famous poem about the sisters – it can find it but it does not present details about it, related things, etc.
• #13: Interesting list from Timothy Burke, Professor of History at Swarthmore, about tools he would like to see when doing research.
• #14: Google’s knowledge graph benefits compared to Tim Berners-Lee’s characteristics of linked data.
• #15: Data sources we work with in OCLC Research. WorldCat is the best proxy we have for the print scholarly record.
• #16:   EntityJS is an OCLC Research prototype for searching across, exploring, and interacting with Linked Data. Using OCLC’s linked data as a starting point, EntityJS provides a way to expand search results to discover not only bibliographic items but also entities associated with bibliographic items such as People, Places, Events, Subjects and Organizations. The application pulls data from a variety of sources such as WorldCat.org, OCLC Works, VIAF, FAST, DBpedia, Wikidata.org and GeoNames. These diverse but connected resources support an entity-oriented discovery experience, a departure from the record-oriented view common in most library systems.   EntityJS is helping us better understand what kinds of data processes and system architectures will work well for RDF Triples ... the basic subject-predicate-object relationships of Linked Data. It provides OCLC researchers with a human-friendly interface for testing discovery and navigation of Linked Data resources, and is part of a testbed for experimenting with user-contributed Linked Data.
• #21: Distant reading – reading at a scale above the individual text. Library-scale. Corpus-scale. ….
• #23: WorldCat identities provides a glancable profile of an author’s publications. Interesting to see what this says about the author’s reception/reputation/etc.
• #25: By aggregating library holdings data, we can see the impact of individual authors in shaping our cultural understanding of ‘place’. Rather than simply counting the publications associated with Dublin and these authors, we computed the aggregate library holdings for their works. This provides a measure of the ‘weight’ of these works and authors in the published record: thousands of libraries around the world have acquired this content. Data based on co-occurrence of author name (VIAF ID) and Dublin geographic heading (FAST ID) in February 2016 WorldCat snapshot.
• #26: We can analyze aggregated library data to identify “important” authors based on their publication record and their relative success in gaining library shelf-share. Here, we have divided the total library holdings for each author by the number of titles (related to Ireland) that they have published. An author with just two best-selling publications will do better in a ranked list like this than a scholar who has published a dozen titles that are not widely collected. So the list is not a measure of the intrinsic merits of individual authors or titles, but a view of how successful they have been in penetrating the global library market. In general, lists like this will favor authors whose works are widely collected by public libraries. This is an important dimension of cultural production and the dissemination of national identity. For many readers in the US, Ireland is genuinely a fiction.
• #30: To identify materials about Ireland, the WorldCat database was scanned to identify all records that contained at least one FAST subject heading that included a string constituting a direct reference to Ireland: i.e., “Ireland” or “Irish”. Headings referencing “Northern Ireland” were filtered out. All FAST subject facets were analyzed: topical, geographic, chronological, personal names, corporate names, events, form, and genre. These headings constituted a core set of Ireland-related subject headings. Any WorldCat record containing one of these subject headings, or a Geographic Area Code of “e-ie” (Ireland), was retained. The next step was to extract all of the FAST headings in WorldCat that co-occurred with the core Ireland-related headings. These were ranked by frequency of occurrence (i.e., number of records in which each heading co-occurred with a core Ireland-related heading). All of the headings that co-occurred ten or more times with one of the core Ireland-related headings were reviewed to assess whether they were also Ireland-related. Given the list of subject headings that survived the review, a second scan of WorldCat was performed to identify all records that contained any of these headings, regardless of whether or not they co-occurred with one of the core Ireland-related headings. The resulting record set was then merged with the first record set produced by the core Ireland-related headings alone, for a total of 417,988 records.   The publications represented by these records account for more than 10 million holdings in library collections worldwide.   Similar analyses done for Scotland and New Zealand indicate that, on average, Ireland-related materials are more widely held than Scottish- or New Zealand-related materials.
• #31: This map shows the largest national concentrations of Ireland-related publications outside of Ireland, as represented by their presence in domestic library collections. The ranking is based on each country’s total holdings of the 417,988 publications identified as being about Ireland. The ranking of the ten largest concentrations correlates with a recent ranking of the most common destinations of the Irish diaspora.  
• #32: The largest institutional collections of Irish-related publications (national libraries excluded). The ranking is based on each institution’s total holdings of the 417,988 publications identified as being about Ireland.
• #33: Most popular works about Ireland and the Irish, ranked by total global library holdings. Each work consists of all related publications that have been explicitly cataloged as being “about Ireland”.
• #34: Word cloud built from the most frequently-occurring subject headings associated with musical recordings about Ireland.
• #35: The largest institutional collections of Irish-related musical recordings (national libraries excluded). The ranking is based on each institution’s total holdings of the 14,922 musical recordings identified as being about Ireland.
• #36: Library data in WorldCat is mostly about decisions that have already been made – acquisitions that were made in the past. But in some cases, the concentration of holdings associated with vendors to the library community tells us something about the continued market interests in a topic. To measure comprehensiveness, we first extract all the library holdings associated with a given author or subject identifier. We then calculate the total number of works associated with the heading, and compute the percentage of works held by individual libraries. Libraries are then ranked according to the breadth of their holdings (their “coverage”) related to the topic or author of interest. More information on the methods behind this work is available here: http://outgoing.typepad.com/outgoing/2013/05/centers-and-coverage.html
• #37: It is important to recognize that our measure of coverage relies on the availability of a FAST or VIAF heading in the bibliographic record. For example, if a library holds works by or about De Valera that have been cataloged in such a way that his name cannot be “controlled” to a VIAF identifier, those holdings will be excluded from the ranked list. Most libraries do not catalog their collections using FAST and VIAF identifiers; OCLC relies on machine processing routines to map local descriptors (names, subjects) to identifiers. Our “coverage” analysis is only possible because of the efforts that are made to organize and enrich the aggregated data.
• #38: Pearse studied for a time at University College Dublin and this may account in part for the library’s relatively high coverage. The strength of coverage at Trinity College Dublin and the National Library likely reflect manuscript and archival holdings in addition to published works. It is important to recognize that institutions that appear to have similar coverage may have very different collections related to Pearse. For example, Boston College may cover a different ~25% of the literature than is available at TCD. In general, however, when we see a ‘plateau’ in coverage across multiple institutions it is an indicator that multiple libraries have acquired the same ‘core’ literature – a microform set, for example, or a series of Collected Works. We make no attempt in this analysis to compare the covered content (titles) across different institutions. Thus the ‘quality’ of coverage may be very different from the numerical advantage of holding a greater share of the literature. One can easily imagine that libraries with relatively low coverage might have exceptionally well curated collections, with a small number of works but a large degree of coherence.
• #39: The bulk of UNC Yeats collection was acquired as a commemorative gift to the library to mark 5 millionth volume. The library’s very strong coverage was achieved in a fell swoop, and not as a deliberate curatorial effort over many years. An inescapable artefact of this approach to computing coverage is the fact that libraries that have catalogued resources as part of a Yeats-related collection (items with Yeats provenance but which are not by or about the poet, for example) will rise to the top.
• #40: Geographic locations of the top ten collections related to four different Irish rebellions. To produce this map, we extracted ‘coverage’ data for four different Irish rebellions using FAST event identifiers. We then compiled the top ten most comprehensive collections for each of the rebellions and plotted their locations based on the libraries physical address. Some locations will have coverage for only one of the rebellions; others will have coverage for multiple rebellions.
• #41: Beyond institutional strengths, one can see regional concentrations of excellence – here the color spectrum of the heat map indicates the average regional ‘coverage’ of the literature. The red zones have a high degree of coverage (at one or more institutions) for literature related to Irish rebellions. This map relies on the same underlying data as the previous ‘location’ map, but it aggregates the coverage into regional averages. The red areas will have on average 40% or more of the related literature; blue areas will have only about 20%. Because the map aggregates data for four different subject areas, the averages are computed across all of the rebellions. So a region that has superior coverage of Emmet’s rebellion but relatively limited coverage of the Easter Rising, will tend to have a yellow profile.
• #42: In New England, there are several repositories that hold substantial collections related to multiple rebellions – from a practical point of view, this could be useful information for planning a research trip.
• #43: In the Midwest, there is a smaller concentration of institutions with broad coverage of the literature, but there is a shared pattern of collecting around the rebellion of 1803. This pattern is revealed by the data – whether or not there is any deliberate coordination of collecting activity at Notre Dame and the University of Illinois, we can see a convergence of library investment reflected in holdings.
• #44: Some regions have broad coverage of literature on multiple rebellions, in multiple libraries. Others have a single institution with exceptionally good coverage of a single rebellion – this is true for Memorial University, which has the best coverage of the Rebellion of 1641.