This news post makes reference to antisemitic language in the context of the research carried out for the DE-BIAS project. Continue reading or go to the Europeana Pro homepage.
Identifying antisemitic terms
The DE-BIAS project aims to promote a more inclusive and respectful approach to the description of digital collections and the telling of stories and histories of minoritised communities. The project covers several sensitive issues, including those related to antisemitic language in Europeana metadata.
As part of the development of a tool to detect this language in the metadata of cultural heritage institutions, DE-BIAS project partners collected and analysed a list of problematic terms related to Jewishness. The terms were collected using a ‘bottom-up approach’. We used words such as ‘Jew’, ‘Jewish’, ‘Jewess’, ‘Israel’, ‘Israelite’ and ‘Hebrew’ as search terms in Europeana.eu to find relevant records, following the methodology of our previous studies (Kizhner et al. 2022, 2023, Zhitomirsky-Geffet and Kizhner 2024). Metadata for these records either contained antisemitic cliches such as ‘Jew Bill’ or a combination of references to the readers’ previous knowledge and an image with a caricature depicting a Jew.
Interpreting such cliches, word combinations or satires requires additional research to understand historical contexts relevant to specific temporalities and societies. The first group of words that we identified are terms denoting antisemitic events, myths and legends, or words connected with these myths and representing an important part of such stories. These are words such as ‘Blood Libel’ referring to an antisemitic myth originating from mediaeval Europe. These are words or word combinations with important consequences leading to persecution, occurrences of auto da fe (trials) in Spain or Portugal, or the events of Shoah in the twentieth century.
The second group of terms is connected with the political rights and perception of Jewry in Britain in the 18th and the 19th century. These are such terms as ‘Jew Bill’ (referring to the Jewish Naturalisation Act of 1753), and ‘Moses Gordon or Wandering Jew’ referring to Lord George Gordon’s conversion to Judaism in 1787. These words are close to the perception of Jews in economic life and street life. Such perceptions are connected to words, such as ‘Jewish moneylender’; bankier, ‘Jewish broker’ or, perhaps, even ‘Hebrew melodies’, a use of parody applied to a street image of a Jew. These terms are usually surrounded by antisemitic textual contexts or antisemitic stereotypical visual representations.
Challenges and complexities in semantics
Our analysis revealed that in addition to straightforward examples of problematic language, there are also more nuanced word combinations which use stereotypes or antisemitic cliches in a less straightforward way. When such words are considered in other contexts, they may shed their antisemitic meaning, creating challenges while working with the semantics of words or word combinations.
For example, Biblical contexts of the Old Testament do not connect the mention of Jews with antisemitism, while mediaeval myths or legends within Europe often assume villainy in Jews. Such terms used in multiple contexts cannot be flagged, using automatic or algorithmic methods. On the other hand, visual representations of Jews, especially young women, may become romantically exotic or oriental, following 19th century fashions, even in Biblical contexts.
Another less straightforward group of terms are such titles as ‘A portrait of an (old) Jew’ where visual representations may range from stereotypical representations of a man with a beard, hooked nose, and a hat to a representation of an exotic oriental personality to Rembrandt’s portraits which include a variety of deep and diverse meanings. If ‘exoticisms’ connected to Jewish portraits may relate to various semantics with a meaning of the ‘other’, such as ‘alien’, ‘oriental’, ‘rough’ or ‘primitive’, there may be portraits which involve other interpretations, not necessarily stereotypical or antisemitic.
Green’s Dictionary of Slang (2010) informs us that, ‘in slang, reflecting centuries of Christian teaching, the Jew is grasping, avaricious, wealthy, untrustworthy, deceitful and mean (as well as circumcised and abstaining from pork). Thus virtually all comb(ination)s with Jew/Jewish are derog(atory), and play on these stereotypes’. However not all contexts which use ‘Jew’ in combinations with other words are necessarily biased. It seems that in popular culture of the 18th - 19th centuries in Britain, indeed, such combinations of words as ‘Jew Bill’ or ‘Jewish broker’ are almost always prone to bias, while in other contexts, such as the descriptions of biblical texts or images, ‘Jew’ and ‘Jewish’ is not connected with derogatory contexts.
Addressing complexities
It is important to acknowledge these challenges because in projects, such as DE-BIAS, the use of automatic tools is restricted to flagging single words or word combinations. When working with text analysis, it may not be within an algorithm's power to detect a biased representation of a man with a beard, hooked nose and a malicious look combined with the word ‘Jew’ in the title or description.
Developing techniques to cope with ambiguous cases where visual representations are combined with textual characteristics is beyond the scope of the DE-BIAS project. However, they can be considered as further steps, and can become the aim of a follow-up project developed by researchers working in digital humanities or major cultural institutions. In doing so, cultural heritage professionals and scholars may add to the knowledge of cultural contexts and improve the understanding of the general public.
The results of this research can also add more data to the indexing and metadata used in the humanities for the analysis of collections of cultural heritage as data. The difficulties of such analysis may include a low number of images with biased representations that can be used to train computer vision models and the subjectivity of defining ‘malicious’ or ‘excoticised’ representation of a Jew.
Find out more
For more information on multimodal analysis and computer vision used for the analysis of cultural data, interested readers may consult the proceedings of Digital Humanities conferences held by the Association of Organisations of Digital Humanities or peer-reviewed journals related to the field of Digital Humanities.
You can also find out more about the DE-BIAS project.