3.3. Format: How can a computer reach this understanding across different types of data?

The availability of digitised and born-digital sources means that we want to ask research questions that draw on more diverse types of evidence. However, our primary and secondary sources can often differ in their format — what might be called mixed data — arising from mixed methods in data collection. 

Traditionally different source types, such as personal diaries and census records, or interview transcripts and survey responses, have to be queried separately because mixed data typically produces different, incompatible data structures and formats. However, an ontology enables us to work across mixed data, irrespective of the sources’ structure and format, because it is concerned with describing entities and how they relate to one another in the real world, rather than describing how the entities are represented within each source’s own structure and format. As such, an ontology can apply this ‘world view’ to data consistently, across all types of sources. 

This is a quality shared by taxonomic approaches, such as those used by NVivo, but with the benefit of bringing meaning through named relationships.

For example, the project Beyond the Multiplex conducted 200 interviews, 16 film elicitation workshops, and a survey of over 5,000 participants. The project was interested in how audiences for specialised film develop. The conventional approach might have been to code the interviews and workshop transcripts so that they could be analysed using NVivo, and then use graphs (SPSS etc) to analyse the nominal and ordinal responses in the surveys. A researcher would then need to analyse and compare each separate set of results. But instead, the project developed an ontology that describes the domain of film and film audiences and then applied this across all the different data types: interviews, workshops, surveys. It meant that the entire dataset could be queried in the same way, and the results could be graphed in the same way, irrespective of each data type’s original structure and format.

So if an interviewee said that she dislikes horror films and three survey respondents responded negatively to the question do you like horror films? we could use the ontology to code four people as disliking the horror film genre.

previous page | next page

URL for this page
https://www.dhi.ac.uk/books/ontology-guide/different-types-of-data