Data analysis, methodological and epistemological issues
Preliminary remark
Humanities subjects often base their mode of knowledge on hermeneutic procedures, in which larger source corpora are described and interpreted in a selective manner.
Although empirical methods (such as oral history or participant observation) are just as familiar to the humanities as statistical methods (for example, in economic and social history), there is often a reserve toward computer-assisted analysis methods, which can be seen in a long tradition of epistemic struggle between quantitative and qualitative approaches.
It must be acknowledged that data analysis methods are more established and enjoy greater acceptance in areas where they concern material, scientifically measurable aspects, such as in archaeology. Similar methods that relate to text and meaning enjoy less acceptance. In addition, modern machine learning methods produce probabilities or plausibilities that often seem unusual as results of qualitative research.
[20231231]
Legal and ethical aspects
The analysis of research data raises issues relating to data protection, copyright, and research ethics. These areas must be weighed against each other in light of the question being addressed, the research objectives, and the publication strategy being pursued.
Social science processing of data is expressly permitted provided that the relevant legal and ethical requirements are complied with. §42h öUrhG regulates the processing of data for data mining purposes.
Legal and ethical aspects of data-oriented research projects must be addressed in data management plans.
[20231231]
Network analysis
Network analysis in the narrower sense refers to the mathematical examination of a structure that arises in a limited corpus/population through the abstraction of entities and relations to nodes and edges. Entities/nodes can be people or institutions as well as words or handwriting, for example; edges/relationships can therefore represent contacts or affiliations as well as joint mentions or transmission contexts.
Since the 1970s, network analysis has offered a professionalized portfolio of calculation methods, originating in the social sciences, through which, for example, the mathematical properties of entire networks (density) or individual edges (centrality) can be determined and compared.
The Network Turn offers a good introduction to the topic for humanities scholars; examples of applied historical network research can be found in the Journal of Historical Network Research.
[20231231]
Topic modeling
In topic modeling, an algorithm runs through a text corpus numerous times to determine which groups of words (“topics”) are most likely to occur together. “Together” refers to predefined subunits (such as individual letters in a letter corpus); the number of words per topic is also determined in advance.
Topic modeling is well suited for the thematic indexing of otherwise unstructured text corpora (e.g., hundreds of volumes of an OCR-processed historical journal). However, the method can also be used for content-related analyses, as in the case of the Encyclopédie or texts on the Socialist International.
The Python library Gensim is often used for topic modeling, as is Mallet, which also offers good tutorials.
[20231231]
Stylometry
As research has known since the 18th century, the style of authors can be determined quantitatively, for example, by their use of rare (or conspicuously absent) words or by morphosyntactic specifics.
As early as the second half of the 20th century, a methodological framework was developed in this area based on open questions of author attribution (Federalist Papers, De imitatione Christi).
A basic overview is provided by Oakes, Statistics for Corpus Linguistics; a recent example deals with the Corpus Platonicum. The R stylo package is frequently used for stylometric analyses
[20231231]
GIS
Geographic Information Systems (or related spatial information systems) generally refer to data processing systems in which spatial coordinates are assigned to data points. Applications include the visualization of humanities subjects on (historical) map material as well as the referencing of archaeological landscape profiles.
An introduction is provided by Geospatial Historian and A Guide to Spatial History. The program QGIS is usually used for historical GIS tasks.
[20231231]