Historian vs. Machine : Testing the Validity of Automated Network Extraction from Inquisitorial Records
| Authors | |
|---|---|
| Year of publication | 2019 |
| Type | Appeared in Conference without Proceedings |
| MU Faculty or unit | |
| Citation | |
| Description | The growing interest in the formal methods of network analysis in historical research has brought into focus older printed resources containing structured data, such as catalogues (of artifacts, inscriptions, manuscripts, etc.), dictionaries (biographical and other), gazetteers, various structured lists, etc. One such type of resource, widely represented in the history of religions as a field, is indices of personal names contained in editions of sources, which may serve for constructing social networks on the basis of the co-occurrence of names in a textual unit (page, manuscript folio, paragraph, etc.). This paper will focus on selected medieval inquisitorial registers as a test case for quantifying the validity of this method through the comparison of networks validated by manual coding with those created automatically on the basis of name co-occurrence. The aim is to show that this seemingly rather rough method can produce comparatively valid outputs if used with discernment and in an iterative way, where the exploration of the outputs helps to fine-tune the initial index of persons and the process of network construction. Another point will be that the validity also depends largely on the kind of research question such an automatically constructed network is used to answer. Finally, the paper will demonstrate that empirical comparison between the outcomes of the manual and the automated approaches is not always in favour of the manual one, as we might expect; in fact, the automated procedure allows some errors typical of human coders to be avoided. |
| Related projects: |