Dividing and Conquering the Streets of Coburg

Carl J. Sciglitano

The records of the Melbourne Social Survey contained in the Wilfred Prest collection in the University Archives are an important resource for researchers studying the living conditions of working-class Melbournians during World War Two. And yet the full contribution of the interviewers who conducted this landmark survey is not immediately apparent from the records themselves. However, by collating the geographical data on the survey forms, we can gain a sense not only of the people involved but also how they organised their work on the survey. What follows is a brief outline of the process I undertook to transcribe and map the surveys of the Coburg municipality using machine learning tools.

Extracting the Information

Machine learning can ease the labour of transcribing historical surveys. I selected 40 surveys from a pool of around 200 to train the Google Document AI service, which uses OCR to convert handwriting on documents into workable database structures. Initially, the training data was manually transcribed to teach the algorithm to interpret the handwriting. Once trained, the AI system processed the remaining records, resulting in 174 transcribed survey records of varying quality.

Machine learning struggles with nuances and contextual understanding. Variations in dates like “21-6-42”, “21st June 42”, and “21/6” needed standardisation for computational processing. Additionally, there were numerous handwritten annotations and corrections which had to be remediated. For example, as shown in Figure 1, the address was identified as “Essendon” even though this had been corrected to “Coburg”.

After correcting obvious mistakes, I geo-coded each survey using the Google Maps API. Under the assumption that all the records were from either Coburg or Pascoe Vale, I confined the geo-coding process to identify addresses that were the nearest match to what was transcribed. For example, as shown in Figure 2, the transcribed “42 Sangood Street” doesn’t exist, but “42 Sargood Street” does. Similarly, the address “2 housedale St” becomes “2 Lonsdale St” and “32 Sharts & Rd” becomes “32 Shorts Rd”. I also accounted for evolving municipal boundaries between Coburg and Pascoe Vale.

The refined dataset was then integrated into a website. Visualising these data on a map revealed additional transcription errors, which required further manual correction.

Divide and Conquer

Most interviews were conducted in June and July 1942, with a small burst in September. The map (Figure 3) shows how the region was divided: interviewer Pat Counihan worked west of Sydney Rd, while M. Warnecke focused east of Sydney Rd and south of Bell St.

Not much is known about M. Warnecke, and even her surname is uncertain. In contrast, Pat Counihan’s history is well-documented due to her marriage to controversial artist Noel Counihan. Pat, originally a teacher in New Zealand, moved to Melbourne with her husband (McNeish 2003, 272). She found the survey work lucrative, earning £5/week, equal to the basic male wage and substantially higher than the gender-scaled salary she received as a teacher (Anderson 2020).

Interviewers were only paid for completed surveys and Pat found that the ‘sight of her wedding ring led women to tell her details of their married lives and to ask a great many questions about health issues and contraception’ (Warne et al 20023, 535). This may have aided her success as an interviewer: she completed 86 surveys in Coburg and Pascoe Vale – almost half of all interviews in the region. Her name also appears on many survey forms in other municipalities.

Drawing firm conclusions from this data is challenging. Some interviewers appear inactive in Coburg but might have been primary interviewers in other areas or occasional helpers. Further research is needed to understand their roles fully. Nonetheless, transcribing and visualising these surveys does shed light on the labour involved and serves as a launching point for further research. The life of Pat Counihan, for example, is one particularly interesting story of a woman who was a tireless, dedicated, independent, and relentless advocate for change, and a vehement guardian of her late husband’s artistic and political contributions.

Carl J. Sciglitano is a PhD Candidate in the History and Philosophy of Science. His thesis explores the ways digital photography shaped the practices and epistemology of astronomers.

References

Anderson, Margaret. (2020). ‘Housewives to Action!’ Old Treasury Building.

McNeish, James. (2003). Dance of the Peacocks: New Zealanders in Exile in the Time of Hitler and Mao Tse-tung. Auckland N.Z.: Random House.

Warne, Ellen, Shurlee Swain, Patricia Grimshaw, and John Lack. (2003). ‘Women in Conversation: A Wartime Social Survey in Melbourne, Australia 1941-43.’ Women’s History Review 12(4): 527–45. https://doi.org/10.1080/09612020300200372.