Putting together the wartime propaganda puzzle

The study of wartime propaganda from the past is increasingly relevant in the light of current events. A large number of sources of World War II propaganda in the
Netherlands has survived. Yet these sources are scattered over isolated silos, residing in fragmentary collections of different media types preserved by different institutions and published on different platforms.

In this presentation, we first discuss the publication of the BNO (Berichtendienst Nederlandsche Omroep) collection, which contains transcripts of broadcasts by the BNO radio broadcaster, a radio news service which spread Nazi propaganda. In particular, we discuss our attempts to date the transcript pages using NER (Named Entity Recognition) and a custom algorithm.

Secondly, we explain how we applied speech recognition (ASR) to wartime radio broadcasts. We discuss the effect of the unique characteristics of the source material on the ASR quality. We explain how the resulting poor quality, in combination with the difficulties in dating the radio transcripts, posed serious challenges for linking radio broadcasts to radio transcripts, which could not be overcome within the scope of this project.

Thirdly, we discuss how the propaganda collections can now be analysed. Despite the lack of links between broadcasts and transcripts, we demonstrate how publishing these and other propaganda collections together in the CLARIAH Media Suite enables the comparison of search results over time across collections, including compensation for imbalances in collection size. We also explain how we used the ASR and OCR results to analyse the textual content of the collections. We present the case of analysing words that frequently occur together with the word ‘Europa’ (Europe) and consider the possibilities and limitations of such analysis given the issues with ASR and OCR quality.

Finally, we conclude by summarising our experiences of the benefits and challenges of using OCR, ASR, NER and custom algorithms to combine propaganda collections.