About the Project
325K990 - Hybrid Artificial Intelligence Models for Historical Text Analysis: A Data-Driven Investigation of Atatürk Era Reader Letters and Government Agenda
Overview
The project examines a large number of reader letters published in Turkey between 1923 and 1938 using AI-based natural language processing (NLP) and large language model (LLM) methods together with a history-from-below approach, aiming to systematically reveal the everyday demands, expectations, and criticisms of the public in Atatürk-era Turkey.
In this period of accelerating modernization, nation-state building, and strongly felt state authority, the project draws not only on reader letters but also on press coverage of the government agenda in contemporary periodicals, in order to grasp the agendas of ordinary people and understand their patterns of interaction with government policy.
The literature on reader letters has so far been shaped largely by limited data; a holistic, data-driven analysis of hundreds of letters has never been carried out, and the use of NLP models adapted to the linguistic characteristics of period Turkish remains very limited. By filling this gap, the project proposes a new conceptual and methodological framework for both the social sciences and humanities and AI research.
The Sentiments of the Turks of İvraca
In a letter we received under the signature of Halil İbrahim from the İvraca district of Bulgaria, the sentiments held by the Turks of İvraca toward the Republic of Turkey are described as follows:
“In our town, the Turks of İvraca have almost entirely taken to wearing hats and caps. The Greek-style fezzes have been cast aside. Everyone cherishes in their heart a great affection for our Republican Government and for His Excellency Gazi Mustafa Kemal Pasha himself, and wishes success upon the path taken by our Republican leadership. Every Turk harbors a profound hatred for the treacherous sultans and fatwa-issuing sheikhs who deceived us until this day, veiling our eyes to a hell and hiding the world from us.
We sincerely congratulate our Republic, which has ordered the casting aside of Greek headwear, the closing of lodges and shrines, and the abolition of the riffraff-lordship, sheikhdom, dervishhood, and discipleship. The twentieth century is not the age of sheikhdom, dervishhood, or discipleship. Today is the age of science, knowledge, and civilization.
Long live the Republic of Turkey! Down with the wretched sultan!”
Cumhuriyet • 10 Temmuz 1932A Two-Way Analysis
Letters and government-agenda coverage are analyzed with natural language processing and large language models, and examined in two directions:
Tracking how decisions taken by the ruling apparatus affected ordinary people.
Observing whether individuals' demands, criticisms, and expectations influenced political, economic, and social policy.
The aim is to reach a holistic historical perspective.
Language Models Adapted to Period Turkish
To capture the linguistic characteristics of reader letters from 1923–1938 — the period's distinctive vocabulary, idioms, and forms of expression — existing language models are adapted through continual learning. This gives general-purpose language models sensitivity to the distinctive structure of historical texts, improving the accuracy and reliability of the analyses, while also offering a historical adaptation framework for Turkish language models.
Data Sources
To hear the voices of broad segments of society, a dataset of periodicals representing different political perspectives was selected: the newspapers Hâkimiyet-i Milliye (Ulus) and Cumhuriyet, close to the government and published throughout almost the entire research period; the newspapers İkdam, Vakit (Kurun), Akşam, and Tanin, addressing different segments of society; and issues of the journals Sebîlürreşâd and İctihad.
Funding and Institutions
TÜBİTAK 3005 — Program for Innovative Solutions in Social Sciences and Humanities Research Projects
The project is funded for a period of 24 months.
The project is carried out with an interdisciplinary approach, bringing together researchers from Pamukkale University (the Department of History in the Faculty of Humanities and Social Sciences, and the Department of Computer Engineering in the Faculty of Engineering) and İzmir Institute of Technology.
Meet the team