WP1 Computer Linguistics

MLU Halle: Prof. Dr. Paul Molitor, Dr. Jörg Ritter, Sascha Heße, Janis Dähne, Dr. Marcus Pöckelmann, Elisabeth Hannah Werner
Univ. Potsdam: Prof. Dr. Manfred Stede, Steffen Frenzel

Development and implementation of innovative algorithms for automatic cross-lingual text comparison at paragraph, sentence, phrase and word level

In this work package, algorithms will be developed that perform alignment between several versions of a text, some of which are in different languages, at different text levels (paragraphs, sentences, phrases, words or elementary discourse units). This is mainly done using word/phrase embeddings, from which semantic similarity can then be calculated. Particular challenges include n-to-m alignments, the quality of the embeddings, the interaction between different levels of text, and the question of efficient algorithms for comparing more than two versions of text. To ensure interoperability, standardised formats for import and export will also be developed in this step.

WP2 Computer Science

MLU Halle: Prof. Dr. Paul Molitor, Dr. Jörg Ritter, Sascha Heße, Janis Dähne, Dr. Marcus Pöckelmann, Elisabeth Hannah Werner

Enhancement of the LERA working environment to include a close-reading visualisation of multilingual alignment results and interaction options for manual post-processing.

An application scenario for the methods developed in WP1 is to provide the generated segmentation and alignment data or variance information in a user-friendly way, so that on the one hand the specialist researchers can check the results of the algorithms and manually edit them if necessary, and on the other hand third parties can browse through the text comparison in an exploratory way. As part of WP2, this will be done as a proof of concept and the LERA work environment (https://lera.uzi.uni-halle.de/) will be extended to include a close-reading visualisation of multilingual alignment results.

WP3 Evaluation and Digital Humanities

FU Berlin: Dr. Brigitte Grote, Sandra Balck, Fabian Etling

Testing and application of (semi-)automatically generated alignment data for selective evaluation, visualisation and use in editorial activities

Work package WP3 is responsible for testing and applying alignment data in selected application scenarios in the field of digital editions. This will be done in close cooperation with the editorial project 'Hannah Arendt. Kritische Gesamtausgabe' at the FU Berlin (https://www.arendteditionprojekt.de). The tasks include investigating the possibilities of selective evaluation and visualisation of (semi-)automatically generated alignment data in the presentation of individual texts. In addition, concepts for the effective use of this data in editorial work will be developed and implemented in order to increase the quality and efficiency of critical editions.

WP4 Digital literary studies 

FU Berlin: Prof. Dr. Frank Fischer, Yashee Singh

Demonstration of the developed technology using prominent cases of literary translations/adaptations as a concrete application of computational literary studies.

The focus of this work package is the exemplary analysis of several prominent translations using the developed tools. The cases come from four time periods (early and late 18th century, 19th and 20th century) and refer to different genres (novel, prose drama, verse drama). Two of the texts have been translated from English into German, the other two in the opposite direction. The corpus is prepared in TEI/XML format and enriched with metadata. By analysing the syntactic, lexical and semantic differences, the aim is to provide insight into the practical use of the tools developed and to validate their applicability for computational literary studies in order to support research processes and promote understanding of multilingual textual adaptations.