Publications
Peer-reviewed academic publications produced by team members within the OKTA and OSPA projects.
- OSPA
Cross-individual sentiment analysis for historical text processing: detecting author-based relationships in Ottoman-Turkish memoirs
Digital Scholarship in the Humanities · Oxford University Press
Authors
Mustafa İlter, Yasemin Özcan Gönülal, Buket Erşahin, Doğan Evecen, Sezen Karabulut, Selma Tekir, Emre Onuç, İbrahim Berci
The authors propose an automated approach to identify sentiments that memoir authors directed toward specific individuals in Late Ottoman–Early Turkish Republic texts. Using a dual-pipeline methodology, they first fine-tuned BERTurk for Named Entity Recognition to detect person mentions. Subsequently, they developed cross-individual sentiment analysis (CISA) with a specialized dual-encoder framework to capture authors' differentiated emotional attitudes toward multiple people in identical passages. Their general sentiment model achieved an F1 score of 0.9262, while the CISA pipeline reached 0.8705, demonstrating potential for analyzing interpersonal connections within extensive historical document collections.
View Article