The digital library in the re-inscription of African cultural heritage

Digital scholarship

Digital libraries in South Africa have seen uneven growth in scale from early grant-funded projects to institutionally based organisational units. The University of Cape Town Libraries has successfully traversed the divide from grant-funded project to internal programme through engaging the academic researcher in critical collaboration through digital scholarship programmes.

While libraries and archives feature significantly in the search for primary research sources, digital humanities researchers increasingly seek to assemble mixed competencies drawn from computer science, design, online publication, digital curation and library science. The digital library is particularly well suited to meeting the needs of digital scholarship by providing technical services across diverse disciplines, and thus facilitating dialogue; furthermore also in promoting ideals such as open access and digital preservation, and championing scholarly and pedagogical innovation. Lisa Shapiro (2011: 1) defines the scope of digital humanities in mediating diverse and sometimes conflicting histories and traditions, rather than technology for the sake of technology:

It can encompass a wide range of work, such as building digital collections, constructing geo-temporal visualizations, analyzing large collections of data, creating 3D models, re-imagining scholarly communication, facilitating participatory scholarship, developing theoretical approaches to the artifacts of digital culture, practicing innovative digital pedagogy, and more.

Thanks to a new emphasis by funders on research data management, the role of the digital library now involves integrating data curation into digital scholarship, enabling a fluency with topics such as repositories, web publication and information-sharing practices, descriptive standards, metadata formats and the plethora of different file formats that characterise digital data. The digital library now encompasses conceptual frameworks for the nature of digital objects, types of metadata, data curation, for understanding collections management as curation, as well as for data sharing and its legal limitations. In transitioning the digital library from a reprographic service to a research partnership, and with the experience of projects such as the Westphal collection of non-Bantu click languages, we perceive the need to develop new collections stewardship policies, new collaborative governance structures and interactive workflows to guide digital librarians and data curators in working alongside researchers in the digital humanities.

In 2011, through the Vice Chancellor’s strategic grant, Humanitec was established to address these divides. The project envisioned the creation of an institution-wide digital repository, to be used to (1) curate and showcase the many important collections of works of art, music, rare documents and other artifacts that are in the possession of the University, and (2) showcase the intellectual and scholarly activities of members of the academic staff. In doing this, the project aimed to connect UCT in important ways with scholars in other countries, by making important African collections available to them for study and by enhancing awareness of areas of research, expertise and collections at UCT.

Voices from the past: A case study of the Westphal collection

UCT Libraries Special Collections and Archives has an outstanding collection focusing on the African continent. Emphasis is placed on African imprints, collected in conjunction with material directly about Africa from other continents. The collection comprises holdings of 85,000 African Studies titles, including also pamphlets, current journals, rare books, an historical collection of maps, archival manuscripts, sound recordings, posters, and African film and photographic records.

The Westphal holdings in the UCT’s Special Collections comprise some 200 audio tapes, handwritten field notes and sketch maps as well as unpublished manuscripts on indigenous languages of Southern Africa. Created by Ernst Westphal, Head of African Languages at UCT between 1962 and 1984, the Westphal sound files are precious because they include recordings of some languages which are no longer spoken and which have not been documented in greater detail.

In 2013–2014 two Humanitec grants supported the digitisation of the audio recordings of Southern Bantu and non-Bantu click languages which Westphal recorded in South Africa, Namibia, Botswana and Angola between the early 1960s and the beginning of the 1980s.

The project of study of the Westphal collections is led by Dr Matthias Brenzinger, Director of CALDi – Centre for African Language Diversity – and Head of Linguistics at UCT. One of his main interests lies in the documentation of African languages, especially those that are vanishing (Brenzinger, 1992, 2007). These studies have become one of the most vibrant fields in linguistics over the past two decades. Research on the actual speech and its variation has regained considerable recognition in academia. While theoretical, universalist approaches to the study of languages had dominated the discipline for almost half a century, scholars working on African languages always had a strong focus on language description and documentation. Linguistic analysis requires a thorough description and a language corpus, which does not exist for most African languages.

Ernst Westphal was a true pioneer in establishing the study of Southern African linguistics and has contributed significantly to our understanding of the non-Bantu click languages. Westphal rejected the term ‘Khoisan’ (Levin, 2015), as he claimed that the non-Bantu click languages spoken by former hunter-gatherers and some of the pastoralists in Southern Africa belong to several unrelated language families. Although currently discredited, Joseph Greenberg’s postulation of the Khoisan language family was widely accepted at the time, while Westphal’s rebuttal was ignored outside of South Africa (Westphal, 1964). Over the last 10 years, however, the findings from research on non-Bantu languages have produced evidence that in fact Westphal was right in rejecting the single family hypothesis for these languages (Güldemann, 2008).

The audio soundtracks digitised in this project include recordings of languages no longer spoken. Westphal recorded the last speakers of Kwadi, //Xegwi and N/amani, and his recordings of N/uu are equally precious as the last handful of remaining three speakers are no longer fully competent in this language.

Westphal’s audio files of Job (Jopi) Mabinda are of special importance. Mabinda spoke //Xegwi, one of the non-Bantu click languages of South Africa and was first recorded by Leonard Lanham and Desmond Hallowes on a farm in the Lake Chrissie district (Mpumalanga Province) in 1954 (Lanham and Hallowes, 1956a, 1956b) Accompanied by Hallowes, Westphal worked with Mabinda in 1962, when the audio tapes in UCT’s Westphal holding were recorded. Oswin Koehler, from the University of Cologne, Germany, visited Mabinda several times between 1967 and 1979, and audio recordings are believed to exist in the Koehler Archive at the University of Frankfurt. Professor Tony Traill, Chair of Linguistics at the University of the Witwatersrand and an authority on non-Bantu click languages, consulted Mabinda in the 1980s and worked with him on //Xegwi until 1988. According to Traill, most //Xegwi community members had abandoned their language and spoke isiZulu as mother tongue as far back as the 1950s. Job Mabinda, however, not only maintained a strong identity as a //Xegwi man, but retained a thorough knowledge of //Xegwi, of which he was the last speaker.

Kwadi used to be spoken at the southern coast of Angola and Westphal’s recordings of the language in the 1970s were with the last, already by then semi-speakers. Very little is known about this language and the data that has now become available through the digitisation project might allow for a better understanding of the history of this language and the genetic relationships of its speakers. In 2014, the German linguist Anne-Maria Fehn conducted fieldwork in the area where Kwadi used to be spoken. When playing audio clips of Kwadi from the UCT Westphal holdings to elderly people who claimed to have Kwadi ancestors, some recalled words from the language they had heard when they were young.

These digitised language recordings not only enable researchers to analyse the language by applying present day technologies and theoretical frameworks, but also allow audio documents recorded some 50 years ago to be returned to family and community members of the recorded speakers. Furthermore the conversion from these reel-to reel tapes to high resolution digital masters ensures the material will remain audible in perpetuity.

Evolution of metadata practice in response to digital scholarship

In 2012, UCT formally approved a Metadata and Information Architecture Policy, jointly owned by the Information and Communications Technology Services and the UCT Libraries. The aim of this policy is to ensure that all content collections, both physical and digital, generated and managed by UCT have sufficient metadata that meets international standards and is consistently applied to ensure that the content is discoverable online. In 2013, the Metadata Working Group was established to implement the policy and provide metadata guidelines and assistance to content collection owners. One of the benefits of strategic institutional support has been the growing awareness across campus of the importance of assigning accurate metadata that meets international standards to content collections. The Humanitec project has afforded the Library, and specifically the librarians in the Cataloguing and Metadata Management Section, the opportunity to work closely with researchers and project owners in metadata creation. As a result, there is now collaboration between researchers, who with their subject expertise can create the metadata that best describes the collections, and librarians who ensure that international standards are applied consistently.

The creation of metadata for the digitised tapes of the Ernst Westphal collection presented unique challenges due to the highly specialised nature of the subject matter. Initial project work preceded the structured workflow initiated by the Humanitec project, as a result of which the creation of the metadata fell to librarians with no subject specialisation and was created from the physical object such as the tape box and not the actual tape. The limitations of this method became very obvious when the opportunity to work with specialists did arise in subsequent project phases.

These later phases were funded, allowing for the employment of linguistic specialists, who were able to listen to the tapes and correctly identify the languages spoken, as well as to follow the subject content of the discussions. In some instances, the identities of Westphal’s interviewees were revealed. Therefore, beyond the essential level of descriptive metadata, these collaborations with researchers enabled further valuable enhancement opportunities.

With specialist input, greater consistency in the spelling of the languages could be achieved, in keeping with internationally accepted taxonomies. This often differed from the terminology used by Westphal to describe the content of the tape boxes. For example, he used the term ŋ|hu for the N|huki language; and Yei and Yeyi interchangeably. Other challenges where subject specialisation was essential lay in the identification of the special characters used by many of the non-Bantu click languages to indicate clicks. Some of these characters are standard Latin ones, for example in ! Xuun, whereas others require the Charis SIL font to provide the additional characters, such as ≠aka-ŋ|ous.

In addition to ensuring consistency in spelling and terminology, metadata librarians provided additional value in adherence to international standards, including Dublin Core (unqualified) and Library of Congress subject headings. Further linguistic standards were also applied, including the codes developed by Malcolm Guthrie for the classification of the Bantu languages (Maho, 2009).

As mentioned above, Westphal himself was not always consistent in his spelling and terminology, and it was considered important to capture these idiosyncrasies in the metadata. Considerable value was added to the resource by capturing annotations reflected on the tape boxes to the description field. For example: Title on box: Kwisi words (Capelopopo).

The benefits of close engagement and collaboration between librarians with their expertise in the application of international standards and researchers with their extensive subject knowledge have resulted in very rich metadata which ensures that these rare and valuable recordings will be discoverable to future researchers.

The challenge of digitisation: Technical service or digital scholarship?

The linguistics scholar Matthias Brenzinger first consulted the Centre for Popular Memory at UCT on the digital preservation of a handful of field recordings from the 1960s on quarter-inch tape reels. At the time, the project presented a novel relief from the systematic digital preservation of several thousands of cassette tapes of oral history interviews. The project was later transferred to the Digitisation and Digital Services Section of the UCT Libraries, as the scholarly engagement was channelled through the Humanitec project.

An early challenge of digitisation was in sourcing a professional reel-to-reel recorder to add to the CPM’s collection of legacy format playback equipment and have it serviced by a rare enthusiast, who proved to be very forward thinking in terms of backwards compatibility, and had been collecting disused equipment for many years. Once the new machine had been installed and a test run had been performed with a non-critical tape, the first originals from the Westphal collection arrived. As alluded to in the previous section, basic metadata had been provided by the Cataloguing and Metadata Section, with an accession number in small pencil lettering on each box and reel.

It was an exciting moment, threading the first end of tape onto the take-up reel, setting the audio software on the computer to start recording, and hitting play on the Revox. Would the old tape tear? What would we hear? Would we be able to capture the full audio spectrum adequately? From the very first tape, the clarity of the sound was astounding. The playback head was perfectly calibrated, with no need for adjusting the azimuth, and the tapes had obviously been stored correctly, showing no signs of stretching, brittleness, mould or excessive dust. But of course there were other interesting challenges. Westphal had recorded in mono, forwards and backwards. He had also quite regularly changed the speed at which he was recording, obviously trading off – whenever necessary – between higher quality (high speed, i.e. 7 ½) and more recording time (low speed, i.e. 3 ¾). In practice, this meant that during playback in stereo, one was listening to a cacophony of voices: some sounding backwards and/or pitched twice too high and fast or too low and slow, mixed with some at the right settings, via the left and right speakers respectively. Nevertheless, in order not to stress the tapes unnecessarily, it was decided to digitise the recordings as they were, capturing both channels at once in 24 bit/96 kHz, focusing only on keeping the signal going through the high-quality Lucid AD-converter trimmed as hot as possible without clipping. Thereafter, it was relatively easy to digitally split the dual-mono file, reverse the separate channels, change their speed and pitch, and export new, unfiltered derivative files. These would be almost indistinguishable from ones that would have necessitated a second recording process, i.e. rewinding the tape completely, turning the spool over, rethreading the tape and replaying it at whichever initial speed the first take suggests. From a digital asset management (DAM) point of view, this way of proceeding also meant that we were creating one single archival (.aif) audio file to match a single physical object in the Special Collections archive. Several more challenging details emerged in the digital scholarship engagement, some of which were purely mechanical, while others pertained to digital workflow and research data management (RDM) issues.

On the mechanical side, two to three tapes that contain more than the usual two (e.g. stereo or dual-mono) channels of audio were encountered. The reason for this, it was surmised, could be that Westphal used a different kind of field recorder (capable of simultaneously recording more than two tracks) to record onto these specific tapes, or it could mean that a misalignment of recording heads led to one channel (that was actually meant to be recorded over) remaining audible. This issue remains to be addressed in future investigation. Another challenge lay in correcting inconsistent speed changes on certain recordings that Westphal made while the battery power supply of the field recorder was deep-draining, resulting in a gradual decrease of recording speed. It proved possible to achieve, but these enhancements on the derivatives were not as perfect as the general filtering and processing that was achievable in order to reduce tape hiss and rumbling wind. Lastly, there were also some reels with up to four separate sections of tape wound onto them, but without these necessarily being torn sections of the same recording. This necessarily led to the creation of distinct archival audio files, which became one of the subjects of consideration in data- and metadata management. In fact the biggest challenge for this collection of audio files was only to become entirely clear later: digitisation, digital storage and metadata creation and enhancement had taken place in four successive phases, leading to small inconsistencies in the file naming convention, for instance in the suffixes of derivatives.

An important part of the scholarly engagement identified by Douglas (2000) was the activity of listening itself, which in the context of audio-visual archiving tends often to be overlooked, between technological considerations on the one hand and academic outputs on the other. While the project did require high-resolution original audio files for linguistic analysis with specialised software, such as Praat, filtered derivatives were also created for maximum ‘naked ear’ usability, such as transcription and translation (Boersma and Weenink, 2001). This process of intense listening to these recordings revealed much more than non-Bantu click sounds – of course unintelligible to a non-expert – but many further impressions emerge. For example, an immensely charming interviewer’s voice (namely Westphal’s), speaking a large variety of languages himself, at times to a second translator in the room; ways of speaking with each other that always involve a rich palette of aural signals that apparently need no translation for any listener: empathic sounds, sounds of understanding, interest and also joy; a whole tape side of what may be young relatives of Westphal’s: two boys making up a radio play in the Queen’s English; documentary excerpts of ritualistic singing and drumming that were clearly just a small window into a musical event that stretched for days and nights; two young women singing what sounds like a nursery rhyme – an intricately interwoven polyphony of such startling beauty that one is almost reduced to tears. Suffice to say: there is always so much else contained in any recording however specific to a discipline its provenance may be – details, backgrounds and ‘ways there’ that reach far beyond the immediate research interests of any academic project. Furthermore, with the passing of time the manners in which we can (and indeed need to) engage with these mediated events also change. In the case of the undeniable affective quality and fascinating layering of the Westphal recordings, one can only wish to provide them with more listeners, at the very least in appreciation of the cultural heritage they represent.

While there is no doubt that a high degree of specialisation is unavoidable in meeting the wide range of systemic demands that are put on digitisation and research data management, it is equally true that the very outputs of this work necessitate the more broadly skilled function of the digital curator. In order to create the kind of digital repositories that the digital library of the future requires, it becomes ever more necessary than in any traditional archival context for researchers to work in close collaboration with digital knowledge brokers: staff who can intelligently engage with their materials in order to provide a variety of relevant accessibility and interoperability solutions. The various digitised media formats can always be further interlinked and enriched. Recently, a small-scale, departmentally driven initiative to make reproduction photographs of Westphal’s handwriting on the tape boxes (and some reels) led to the creation of an auxiliary archive of visual material to complement the relevant audio files and text documents. Further efforts could include time-based anchoring of relevant transcripts, meta tags, commentaries and other hyperlinks on the audio files, perhaps via an interactive web-interface with specific access rights put in place for research experts and general audiences. In this sense, the future potential for the Westphal collection is to open it up to new audiences outside of its immediate field of specialisation, and to enable new ways of listening that will no doubt generate interdisciplinary insights into what its richly layered materials provide, in this case particularly to the African Humanities.