We have joined forces in CRETA e.V. because we are working together - although spread over different locations - to develop, investigate, describe and anchor reflected practices of text analysis in the humanities and social sciences. In doing so, we face three major and recurring challenges:
- the operationalization of questions in the humanities and related concepts, i.e. the formal and, if necessary, technical implementation and measurement of the concepts. This includes questions of operational definition and manual annotation or annotation capability as well as questions of automatic recognition. Many questions arise, for example, in the area of quality assurance or validation of operationalizations, which would also allow the comparison of operationalizations. Here, generally accepted strategies, which are appropriate for the objects of investigation, are still missing. The re-integration of quantitative results into hermeneutic workflows and work processes also regularly creates new challenges that need to be addressed.
- the problem-adequate use of the technical possibilities: For machine-supported text analysis, a broad technical inventory of possibilities can be used - from rule-based systems to artificial neural networks. In addition to purely technical performance requirements, however, there are others in the digital humanities that cannot be directly mapped to evaluation metrics. Transparency/interpretability of the models can help to identify ‘blind spots’ of the models, but is also central for a hermeneutic integration. Also the integration of existing knowledge about the domain into the models is in many cases a desideratum.
- the effects of the use of mechanical procedures for the research process in the humanities on the people who conduct it: Even if computers contribute to the gain in knowledge, it remains people who gain the knowledge, formulate and also communicate it. The third challenge is therefore interdisciplinarity as such, meaning both the cooperation between D and H disciplines and the development of solutions across different H disciplines. The question of what the standard procedures or tools of a ‘mature text scholar’ actually are also plays a role here, because their answers help decide which procedures still need to be explained and which do not.
The members of CRETA are working on these three major challenges posed by the use of machine processes in the humanities and social sciences in various concrete projects.
Tracking Character Knowledge