Methodological issues and opportunities

1. As our research project aims at comparing sports coverage on the television and in the newspapers, our first step was to get an idea of the nature and extent of the material available, while simultaneously getting acquainted with the search functionalities of the media suite. One problem to study the remediation of sports coverage turned out to be that while the selected newspapers have been archived integrally and digitized in full, the remaining television footage of sports matches turned out to be very scarce. This meant that when we found potentially interesting articles in the papers, we were often lacking a reference point in the audiovisual collections, even when we limited our research design to football - arguably, the most covered sport in the Netherlands. Moreover, when footage from football matches was found, it was usually limited to just the visual content of the match itself without any of the accompanying commentary. Moreover, the introduction of the coverage by a presenter in the studio and all other content from the broadcast was mostly missing. One explanation for this is that the studio presentation and the voice-over commentary were done live and have not been recorded. An alternative explanation is that these programs did get taped, but the footage got lost due to the reuse of costly film tape. In any case, the context in which football matches were discussed and appropriated is lost - which is a major disadvantage for research into remediation.

2. Next to working with differences in the availability and degree of preservation of archival material, researchers need to deal with the important and time-consuming issue of the pollution of search results. This can imply multiple copies of the same article in different newspapers that pop up in your search results, or ‘false positive’ results that do contain a search term, but are actually not relevant for a researcher’s project. Adding additional search features as well as more metadata would be useful to increase the likelihood of retrieving relevant results. One fairly easy improvement would be linking dates to an eternal calendar, which would enable searching for content that was published or broadcasted only on specific days of the week. News reporting has a certain rhythm. For example, sports coverage will often be published on Mondays, while book reviews commonly appear on Fridays, and specific television programs have a dedicated air time. Another improvement would be to add the genre of newspaper articles and television programs to the metadata - preferably as part of a uniform and consistent categorization. It would be convenient for researchers to only search in book reviews, or in parliamentary reports, news reels or talk shows. Another feature helps to quickly scan the results: the interface of newspaper archive Delpher shows three sentences close to a keyword in the results list. This gives the researcher the opportunity to determine on that basis whether it is not interesting at all or still worth clicking on.

3. Another related issue pertains to the diverging search options for digitized newspaper archives and the audiovisual collections, which differ considerably. While newspapers have only recently been digitized, in longitudinal batches that run from the start of the paper to either its closure or – due to copyright laws – 1995, they have uniform full text search functions through keyword search. This is very different from the audiovisual archives which are searchable mostly through metadata. These descriptions of the content are often very brief and sometimes incomplete, making it much harder to find material than in the newspaper archives.

The most important effect of only being able to search the metadata is that one can only search on a far more general level, limiting the search options considerably. One way to improve searchability, which archives are actively pursuing, is using speech recognition techniques to transcribe spoken text in audiovisual content. To a certain extent, this enables similar possibilities as a full text search. Yet, even when spoken text is transcribed, the options remain far more limited than searching full text in the newspaper archives because the context of the spoken text is missing. Therefore, results are harder to interpret. And of course, the visual features, which in themselves also provide much information, are still not searchable. Finally, given these considerable differences in search possibilities, it is also hard to present search results from both collections in one interface and in a meaningful and comparable way.

4. Apart from the generally limited search possibilities of metadata compared to full-text search, the consistently differing nature and quality of metadata between collections also poses problems. While in the newspaper archives the metadata that have been added are limited to the kind of article (advertisement, personal announcements, illustration, and editorial articles), distribution area (national, local, or one of the former Dutch colonies) and place of publication, these are post hoc allocated in a logical, uniform and consistent way. In contrast, the audiovisual archives have a long historical genesis with different curatorial strategies and systems of categorizing material. This means that, depending on the moment the material was added to the archive, different systems for assigning metadata with their own set of terms and categories, have been applied. This makes the results of search queries rather unreliable, and it requires quite a bit of knowledge about the structure and development of the archive to get (all) relevant results - far more than is necessary for searching newspaper archives. This also implies that while a researcher can be quite confident to get all results when doing a search in the latter archive, they can be less sure about the hits generated in the audiovisual collection. This makes a quantitative exploration of the data - for example by listing numbers of articles and television items on a timeline - a rather treacherous, if not impossible, endeavor which is hard to substantiate methodologically.

5. Storage functions are very important for an investigator to keep an overview and to get back to relevant articles and videos at a later stage of the investigation. The Media Suite offers good possibilities for this. Newspaper pages can be saved in a list of favorites and downloaded, and articles can be cut from the page and downloaded as a jpg-file. Moreover, the permalink and a complete reference are easy to copy. Similarly, different items can be cut from the televisions program they were part of and stored. It is convenient to be able to create different personal folders in the interface so articles and items can be orderly stored and easily retrieved. Furthermore, after (or during) the compilation and storage of research material, the Clariah Media Suite provides (video and textual) annotation tools, which make it possible for researchers to add metadata themselves. This helps them to keep track of their results, but also to analyze the data according to specific research questions. For audiovisual archives, to determine whether a program or item is interesting for research, it is useful when the metadata contains a minute-to-minute description of the content. The catalogue of the Netherlands Institute for Sound and Vision, for example, helps the researcher to quickly slide to a certain moment in the broadcast. If the detailed descriptions of the material would be displayed clearly (preferably next to the video), this would be a big plus for viewing the visual material.

6. In order to compare television and newspapers both databases must be accessible in one interface. In the Media Suite and in a previous tool, AV Researcher XL, newspapers and audiovisual material can be searched simultaneously and plotted on a graph that visualizes the number of hits in a certain timeframe. This makes it much easier for the researcher to match and compare relevant hits. Yet, such a comparative search raises a serious issue in terms of segmentation. In newspaper archives, articles are sometimes incorrectly segmented on the page, i.e. multiple articles are considered as one, or one article has been split in various segments that are considered separate pieces. In contrast, in the audiovisual archive whole television programs count as separate units in the search results. In terms of the newspaper, this would mean that the entire page or even the entire newspaper would be included in the hits instead of an article. Similar to searching in newspapers, as a researcher you would prefer to search and count different items in a broadcast. For example, if a newsreel consists of ten different items about different topics, you would like to be able to search, count, and analyze only the ones you are interested in. This structural difference between the level on which the content in newspaper and audiovisual archives is segmented results in a severe impediment of making a systematic comparison between, in our case, the development of sports coverage on television and in the newspaper. In order not to compare apples with oranges, it is important for a comparative analysis between newspapers and television to equalize the measurement level between both collections as much as possible. This means that segmentation must also take place at item level in audiovisual collections.

Final Remarks

Our research pilot took place in the initial phase of developing the CLARIAH Media Suite. Developing such a complex research environment comes with challenges. One challenge such an ambitious endeavor faces is copyright issues related to the collections involved. It is incredibly challenging for individual archival institutions to reach agreements with publishers, broadcasters and media makers such as journalists and photographers to make collections available either on-site or through their website. Bringing them all together in an encompassing research environment and making them searchable in one interface, brings up new issues that need to be solved both legally and technically (for instance by approaching collections from a central research environment but keep them stored on the servers of the individual archival institutions). Solving such issues takes time, which meant that not all of these issues were solved yet during our project. This meant that comparative search could not be implemented yet in the Media Suite, so we did our pilot research through parallel searches in the individual collections, and by using a previous version of the comparative tool on a much smaller part of the collections.

A major benefit from the Dutch situation compared to other countries, however, is that newspaper and audiovisual collections are quite complete at the Royal Library and the Netherlands Institute for Sound and Vision respectively. While in other countries especially newspapers are part of different public and commercial archives and settings - which makes it even more complicated to link them in a research infrastructure - the Dutch collections are mostly publicly available at a limited number of institutions, which makes it easier to link them and accomplish comparative search options. When this is achieved and the research tools within the Media Suite as well as opportunities to work with the raw data are added, fascinating opportunities for doing intermedia and comparative research are opened up that allow for research on a far larger scale and deeper level than is currently possible.⁸

What is crucial, though, is keeping in mind that in the construction of such an extensive research infrastructure making collections digitally accessible and (simultaneously) exploring (various) collections is not the ultimate goal. While enabling digital access and facilitating various search opportunities is important in itself - and a major leap forward in comparison with what was possible before digitizing collections - the next step in media and digital humanities research should be to use the infrastructure to do (digital) research that answers pressing research questions in the field of media history and media studies at large. During our pilot project the Clariah Media Suite, and the comparative search recipe in particular, were still under development. But working with the infrastructure and the individual digital archives has shown the potential for further grounding remediation as a theoretical concept and answering research questions related to it that were earlier hard to study empirically.