A short introduction to the Jean Desmet Collection and the MIMEHIST project
The main goal of the project ‘MIMEHIST: Annotating Eye’s Jean Desmet Collection’ (University of Amsterdam, 2017–2018) was to unlock Eye Filmmuseum’s Jean Desmet Collection in the CLARIAH Media Suite by making the Collection’s digitised materials available within the environment and enriching them in order to make them more searchable. The Jean Desmet Collection contains the archives of film distributor and cinema owner Jean Desmet (1875-1956) who acquired and distributed a significant amount of films from several European countries and the US made in the early period of silent cinema and its transitional years from the mid-1900s to the mid-1910s. The Desmet Collection is unique because of its large amount of rare films from different countries – containing numerous films that have not been preserved in their country of production – and because of the richness of its business archive which holds extensive documentation of early film exhibition and distribution practices in the 1910s. These features contribute to its immense historical value and international reputation and are also the main reasons why the collection was inscribed on UNESCO’s Memory of the World Register in 2011. The collection consists of approximately 950 films produced between 1907 and 1916, a paper archive consisting of around 127,000 documents that contains both personal and business correspondences (Desmet meticulously preserved most of his transactions ranging from postcards to tax bills), some 1050 film posters and around 1500 photos relating both to films in Desmet’s distribution catalogue and to his personal life. Parts of the collection were acquired by Eye Filmmuseum shortly after Desmet’s death in 1957 and then gradually expanded throughout the years with additional acquisitions.1
In unlocking the collection, we were confronted with two specific problems. On the one hand, it is a materially heterogeneous and complex collection that comprises a large variety of document types that are not easy to make searchable. Especially the paper collection is complex to describe because the items in it often contain images, text and graphic design elements that are sometimes equally important. This combined with the sheer number of items in this subcollection makes the task of describing the documents in it at item level too large a task for Eye Filmmuseum to take on with the resources it has available. In order to solve this problem in the MIMEHIST project, we experimented with a combination of data-driven approaches, using text mining for analysis of typewritten (parts of) documents and computer vision for image and handwriting recognition. As an outcome of this approach we were able to make a large part of the Desmet paper archive searchable through OCR while also offering a proof of concept of how the collection could be browsed in the future based on visual features. The intricate methodological and epistemological implications of this endeavour has been discussed in great detail in an article published in 2018 in a special issue of this journal on ‘Big Data Histories’.2
The second problem we were confronted with was that it had previously only to a very limited extent been possible for researchers to establish and explore meaningful links between the collection’s varying materials. The Filmmuseum observed that even after the materials’ digitisation, scholars would still tend to study the various subcollections separately from each other, even while there were meaningful connections to be explored. For instance, Desmet’s business documents can greatly enrich scholarly understandings of how films – both preserved and lost – were distributed and programmed, and the collection’s posters may tell a great deal about the promotional strategies associated with the films. In order to bring the digitised collection materials closer together to nurture the exploration of links between them, MIMEHIST considered content-driven, video annotation tools that allow for linking video and digitised paper materials a particularly elegant solution. As a complement to the article publishedin TMG Journal for Media History in 2018, this short article expands that piece’s discussion of how MIMEHIST took inspiration from previous decades’ multimedia and video annotation scholarship in film studies to conceptualise video annotation functionalities in the Media Suite. The text first discusses how MIMEHIST situated itself in relation to past and current video annotation and editing approaches. Subsequently, a few examples are offered of how the Desmet Collection could be studied in the Media Suite using such approaches, as a way to point to its future potential for scholars and students.
Scholarly background: MIMEHIST, multimedia editions and video annotation
In thinking about how to facilitate the combined study of the Desmet Collection’s various materials with video annotation tools, MIMEHIST took inspiration from the past decades’ pioneering scholarly efforts to develop multimedia editions for historical contextualisation of (archival) film. In particular, the project built on methods developed for scholarly CD-ROM and DVD editions of archival films, and current research in the area of what is known as hypervideo, that covers practices of annotating and presenting digital video with links to contextual materials such as text, paper and image items. This section discusses a few core concerns of such scholarship illustrated with key examples.
To begin with, one may say that on a general level, video annotation scholarship in film studies has sought to fulfil a wish that has existed for decades, to engage more fully with film in research and teaching, in formats other than writing. For instance, in his article from 1975, “From a Written Film History to a Visual Film History”, film scholar and co-founder of the Harvard Film Archive, Vlada Petric, insisted that the holdings of film archives should be more present in historical scholarship and yearned for a future that would allow scholars more easily to scrutinise film prints as the basis for a ‘visual-analytical’ approach to film history. As Petric wrote:
(…) the appropriate methodology of film history cannot be attained in our time without the full cooperation of the film archives, which possess the prints and have access to technical facilities, without which it is impossible to grasp the cinematic structure of a film.3
Reflecting such a wish, the 1990s and 2000s saw advanced efforts to integrate audio-visual materials more fully into film history teaching and research, through the development of hypermedia formats, that allowed for studying archival films and related materials in combination.4 These formats allowed for commenting on the content of films – for instance their style or forms of representation – and their contextual aspects, such as distribution, promotion and exhibition. In the 1990s, several ground-breaking projects were developed such as Lauren Rabinovitz’ The Rebecca Project (1995), the multimedia textbook The Virtual Screening Room developed at MIT by Henry Jenkins, Ben Singer, Ellen Draper and Janet Murray between 1992-1999, as well as Yuri Tsivian’s CD-ROM on pre-Soviet silent cinema Immaterial Bodies: Cultural Anatomy of Early Russian Films (2000) released in the University of Southern California’s Labyrinth project’s Cine-Discs series edited by Marsha Kinder.
The 2000s’ scholarship on video annotation extended these pursuits by seeking to merge text edition and philological theory with the DVD format’s information architecture. An important goal for this type of scholarship was to create annotated versions of archival films inspired by literary studies’ enhanced electronic text editions and text-critical principles. This should allow scholars to analyse digitised archival films against the backdrop of their historical context by making use of shot segmentation, annotation and hyperlinks to film-related materials. Following philological principles, DVDs were developed which used on-screen footnotes to relate textual features to contextual sources, containing carefully conceived ‘tables of contents, indexes and direct links from each scene to accompanying materials.’5 In film and media studies these methods were not only particularly useful for content analysis but also for, for instance, the critical study of multiple language versions and different cuts of films.
A particularly inspirational example of such scholarship for MIMEHIST is the DVD series of annotated Russian and Soviet classics Hyperkino/KinoAcademia launched in 2008 (Absolutmedien and RUSCICO), developed by film scholars and archivists Natascha Drubek and Nikolai Izvolov. Hyperkino allowed for exploring film as a hypertextual ‘networked index’ or, as suggested by Drubek and Izvolov, a form of ‘hyperkino’, which enables navigation between textual segments and contextual material, for instance paper clippings, production documents, stills or censorship reports. The editions’ footnotes – created by invited film historians – commented on for instance a director’s distinct style, a film’s distribution, promotion, censorship history and archival life. In the releases, annotations appear as clickable numbers on-screen, which provide access to the footnotes (see Figure 1).
To give an example of how such a format can support film historical scholarship one may consider the Hyperkino edition of Engineer Prite’s Project (Lev Kuleshov, 1918), a film considered to have played a crucial role in the Soviet montage film’s development of editing and mise-en-scène. Containing content-specific comments and links to film-related material, the edition’s annotations highlight how Engineer Prite’s Project introduced rapid editing, close-ups and stylistic elements into Russian filmmaking, inspired by popular American and Scandinavian cinema. Beyond stylistic observations, the footnotes also engage in a comparison of the film’s different archival versions to hypothesise on the plausibility of different reconstructions while reflecting on the inherent uncertainties of archival elements and film versions that often have missing segments. For instance, the edition’s footnote ‘The film’s archive life and reconstruction’ discusses doubts concerning the accuracy of the film’s current montage, by referring to contextual materials and existing archival elements, illustrated with a scan of visible editing marks in one of the elements used for the film’s restoration (see Figure 2).
Inspired by the hypermedia projects and approaches discussed above, MIMEHIST in several ways sought to facilitate a research situation that allows for the use of video annotation tools as a way to link and comment on the Desmet Collection’s materials. For instance, the Media Suite’s environment allows for creating an annotated version of a film title in a way that reminisces a Hyperkino approach. One may contextualise a title’s content, production, distribution and archival life by linking films and related documents and synthesise the main findings in annotations in the Media Suite’s personal workspace environment which is the area of the research environment in which items can be bookmarked, compared and linked. In the MIMEHIST project, we tried to do this for a few titles in the Desmet Collection that are known to be incomplete, based on a list created with help from Elif Rongen-Kaynakçi, Curator Silent Film at Eye Filmmuseum. The list comprises titles such as Dødsspring til Hest fra Cirkuskuplen (The Great Circus Catastrophe, Denmark, 1912), When the Earth Trembled (USA, 1913), L’enfant de Paris (France, 1913) and Tragico Convegno (Italy, 1915). As interesting titles on their own, one may use the Media Suite to carry out a content analysis of the films, while also using the Media Suite’s possibilities to search and browse business documents and promotional materials to trace their acquisition and distribution, and look for possible reasons as to why they are incomplete.
In my own research, I focussed on the documents relating to the Danish silent film Dødsspring til Hest fra Cirkuskuplen. A sensational title produced by the then dominating production company Nordisk, and one of its biggest selling titles at the time, the Desmet Collection holds one of the approximately 245 copies estimated to be sold by the company.6 While searching for documents that could reveal why the film was left in an incomplete state, the Media Suite enabled me to establish interesting circumstances of its acquisition and distribution – for instance that Desmet acquired the film via Nordisk’s Berlin offices and was able to premiere it only a few days after the film’s Danish premiere – and gather this in connection to the film title in the same work environment. In doing so, I could begin to develop a sense of the film’s material life and history before it entered Eye Filmmuseum’s archive though not finding an answer as to why it was incomplete through the sources I came across throughout the project.
As part of the project, we also went beyond these practices to facilitate the inclusion of a greater variety of analytical approaches. While the practices established in the 1990s and 2000s were ground-breaking in how they combined philological theory and edition theory with film and media history perspectives to create digital editions of archival films, the closed nature of the physical disc formats they relied on also meant they were limited to one specific historical interpretation – usually limited to the perspective of one scholar – of one or a few films at a time. To challenge this limitation, film and media scholars in the 2010s began going beyond physical disc formats to develop and work with video annotation software that held the promise of being more inclusive of a greater variety of personal and subjective perspectives and interpretations of archival films. For instance, the Pompidou Centre’s Research & Development department developed the video annotation software Lignes de temps – involving in the process film scholars Raymond Bellour and Jean-Louis Comolli to think critically about the affordances of video annotation from an apparatus theoretical perspective. Once developed, the Lignes de temps software allowed scholars to upload, segment, re-edit and annotate digitised films and related materials of their own choice using their own vocabulary, in a desktop environment (see Figure 3).
Beyond film and media studies, MIMEHIST also considered qualitative methodology and grounded theory approaches in the social sciences, anthropology and ethnography that may be said to resemble what Lignes de Temps aspired to achieve.7 Scholars in these fields often use video and audio annotation software to tag and label interview and content bits and organise them into different categories and color-coded layers on a timeline, while linking them to related materials. This analytical process lasts until the researcher feels the material is exhausted or ‘saturated’ in a subjective – or collaborative – coding process, without a single logical, predictable end point.8
Taking inspiration from such video annotation practices, the Media Suite offers students and scholars the possibility to devise their own coding schemes in order to organise and analyse the Desmet Collection’s materials, following their own individual, historiographical and analytical interests, while working in the environment’s personal workspaces. Hopefully, this may allow for a greater multiplicity of theoretical angles to be explored in relation to the Desmet Collection’s materials which we may not have been able to foresee and, by the same token, nurture a more open-ended and dynamic historical interpretation of them, that allow scholars to make different – perhaps competing – annotated versions of the same film titles. A concrete benefit of this may be that researchers and students that work with the Media Suite may challenge existing scholarly and archival interpretations of the materials in their own ways. After the completion of the MIMEHIST project, this endeavour has been carried on in the context of the Media Suite Learn initiative, that aims to integrate the environment into university-level teaching and research through the development of tutorials and the organisation of workshops. In this context, tutorials based on the Desmet Collection have been developed that introduce students to principles of film historical analysis and analysis of (early) cinema’s promotion and distribution practices.9
With the MIMEHIST project we have made a great step forward in unlocking the digitised materials in the Desmet Collection in a scholarly, online research environment. The Media Suite’s environment makes it possible for a much wider group of students and scholars in the Netherlands than hitherto, to research the Collection’s materials and explore links between them. Drawing on both film studies approaches to video annotation and on developments in other fields, MIMEHIST has contributed to creating a work environment that supports established historical methodologies, while also inviting scholars to bring their own theoretical perspectives into the Media Suite by presenting analyses and interpretations that the Collection’s archival metadata and previous scholarship have not taken into consideration. Finally, on a more general level, one may also return to Petric’s plea for closer collaboration between film scholars and film archives made in the mid-1970s, and consider MIMEHIST’s result a contribution to this endeavour that shows how such a collaboration may be configured today.