Daidalos

Development of an infrastructure for the use of natural language processing (NLP) for researchers in the field of Classics

Classical philology is undergoing a process of digitisation. Until a few decades ago, researchers had no choice but to make their way through their texts by hand. Although this way of working produced impressive and exemplary results, the digital research tools available today have the potential to save time and minimise the risk of errors. Daidalos is by no means the first Classics project to focus on digital research methods. Its unique selling point is that it attempts to combine existing digital methods of text analysis and visualisation in a web-based research infrastructure. It also seeks to evaluate and further develop these methods in an evidence-based and easily comprehensible way. Daidalos aims to be a platform for interested researchers (community of practice) and a learning environment where researchers can develop and expand their digital skills (digital literacy) with the help of various autodidactic learning modules.

Tailoring to needs of users through research teams (tandems)

To make an informed selection of the text corpora, analysis tools, methods of visualisation and learning modules offered by Daidalos, the first step for the project team is to evaluate the potential of selected NLP methods. This is done on the basis of real research questions (e.g. on authorship, intertextuality or discourse strategies) that are presented to the Daidalos project by members of the research community. Collaboration in these research tandems is organised as follows: In an initial consultation, the researchers present their research interest, a possible research question and the desired text corpus. The Daidalos team advises them on the specifics of the research question and explains which analysis tools (e.g. morphosyntactic taggers, working with word embeddings) and visualisation options (e.g. annotated text, graph) might be suitable. The machine language processing tools offered by Daidalos make it possible to analyse both linguistic phenomena (e.g. word types, collocations, syntax) and literary questions (e.g. genre features, paraphrase search, ancient concepts). As soon as the results of the analyses are available, the Daidalos team hands them over to the researchers for assessment. In a follow-up meeting, the researchers evaluate the quality of the results and the visualisation together with the Daidalos team. If necessary, the parameters are adjusted and further analyses are carried out. The insights gained from these use cases are channelled into the development of the website infrastructure and the design of the training material.

Interactive research infrastructure as a learning offer

The interactive access to ancient texts is not only intended to help answer research questions, but also to promote methodological reflection and empower users. Accordingly, the software offers the following content:
(a) Demo-workflows with configurable elements, e.g. corpus selection, with detailed explanations;
(b) a domain-specific interface with various features including interactions on research interest and method selection; storage of personalised settings; exercises and code snippets in Jupyter Notebooks;
(c) free configuration of text and method triangulation, integration of own source code for experienced users.