The TurkuNLP research group at the University of Turku invites applications for 2–4 fixed-term postdoctoral researcher positions in digital linguistics and Natural Language Processing (NLP). Employment begins in March / April 2026 at the earliest (or upon agreement) and continues for 24 months or until the end of May 2028.
TurkuNLP is a multidisciplinary research group combining digital linguistics, corpus linguistics, NLP, and machine learning. We specialize in developing NLP tools and large language models (LLMs) for Finnish and other languages, as well as in applying these methods to state-of-the-art corpus linguistics and digital humanities research.
In recent years, a central focus of our research has been language use and its modeling in massively multilingual, web-scale data. In particular, we study web registers (genres) and develop methods for their automatic identification and analysis. TurkuNLP has developed FinBERT and FinGPT, Finland’s leading BERT- and GPT-3–based language models, and we have recently demonstrated that register-based sampling from web data significantly improves LLM training (Myntti et al. 2025).
The postdoctoral positions are connected to three closely related research projects, offering opportunities to work at the intersection of theoretical questions about language use and its variation as well as large-scale computational modeling.
The project “Mechanisms of register variation in massively multilingual web-scale corpora” (Mecha), funded by the Research Council of Finland, investigates language use in the digital world at a truly multilingual scale. The project focuses on web registers such as news, blogs, and how-to pages. Its core goals are to analyze the linguistic characteristics of registers across languages and to understand how and why both the inventory of web registers and their linguistic properties vary—not only across languages, but also across countries and geographic regions. A central research question concerns the mechanisms driving this variation.
Fin-Clariah is a national research infrastructure for the Social Sciences and Humanities, involving partners from most Finnish universities. TurkuNLP’s contribution focuses on modeling noisy web data, with particular emphasis on social media texts. The data and methods largely overlap with those used in Mecha, and a key aim is to transfer methodological advances and empirical insights from Mecha and earlier projects into sustainable infrastructure that supports a wide range of SSH research.
The third project is part of the newly elected Centre of Excellence for Human Diversity Through Contacts at the University of Turku. The Centre studies human diversity from multiple perspectives, including language evolution, genetics, health, and web language use. Within this framework, the project led by TurkuNLP focuses on language use in massively multilingual web datasets, targeting in particular its variation across languages, countries and levels of language contact. Contacts are investigated in particular (1) through similarities and differences in registers across languages, countries, and geographic regions, and (2) through the circulation of attitudes, ideologies, and ideas within and across registers.
Across all projects, we apply and develop state-of-the-art machine learning methods in combination with large, multilingual web-crawled datasets. Beyond methodological advances and linguistic insights, the projects aim to produce high-quality metadata for web datasets, enabling their more informed and responsible use in future research.
For more information, please visit turkunlp.org and consult our recent publications.
Job description
Through this call, we aim to establish a multidisciplinary team of researchers with complementary skills in corpus linguistics, digital linguistics, computational linguistics and NLP, working together towards the goals of the three projects as well as advancing research in their respective fields.
The team will be integrated within the TurkuNLP research environment and the researchers will therefore be able to utilize and further develop the datasets and methodological approaches already established within TurkuNLP. However, the specific approaches and research questions will depend on the candidates' profiles and interests. Possible topics include, but are not limited to:
Modeling and analyzing language use across varieties such as registers (genres) in massively multilingual web-scale data, using multilingual deep learning methodology, LLMs and corpus linguistic statistical methods
- Analyzing the variation of language use (registers, genres) across factors such as language, country, geographic region and level of language contact? What types of characteristics are shared, and which are not? How does the set of web registers and their characteristics vary and why? What are the mechanisms driving this variation?
- Modeling attitudes, ideologies and ideas in web-scale data? How are they repeated within and across different registers, languages, countries and geographic regions? What are the mechanisms driving this repetition?
Developing NLP methods for advanced semantic analysis and labeling of multilingual web data, such as:
- Metadata-informed LLM training based on parameters such as register, ideology, language, culture.
- Data-driven similarity and information gap analyses on large-scale, massively multilingual web data to identify cross-lingual similarities (e.g. topical and register coverage), and detect missing or underrepresented information across languages.
- Efficient, controlled multilingual embedding models for large-scale similarity comparison, especially targeting to explicitly capture different aspects of similarity, such as topical similarity, stylistic similarity, language-related similarity, or their combination.
Theoretical and practical advances for web language use
- Geolocating web language use: developing methods to identify and analyze geographic variation in web registers (comparing language with geography)
- Developing more efficient and theoretically motivated ways to model linguistic variation
- Developing new models of register variation that move beyond categorical models toward continuous representations of situational and linguistic variation
- Exploring and comparing theoretical approaches to register: e.g., investigating how Systemic Functional Linguistics (SFL) could be integrated with corpus-driven register analysis. Examining whether SFL's theoretically motivated parameters (field, tenor, mode) yield more principled or cross-linguistically valid register distinctions than other corpus-linguistic approaches.
Additionally, the candidate is expected to participate in the supervision of junior researchers and planning of funding proposals.
The postdoc's work tasks also include some teaching (e.g., 5% of working time or upon agreement).
Who we are looking for
The project goals span both linguistics and NLP, and candidates with various backgrounds in, e.g., corpus linguistics, computer science, or related fields are encouraged to apply. We hope you have
- An appropriate doctoral degree
- Programming skills (python / R, command-line Unix environments (CSC))
- Knowledge of computational methods in NLP and/or corpus linguistics (such as deep learning, Huggingface, processing very large datasets, statistical modeling)
- Depending on your profile, familiarity with register studies, corpus linguistics, and/or machine learning methods in NLP
- Ability to work independently while also taking responsibility for the research group's activities
- An understanding of multidisciplinary research
We primarily seek to recruit a Postdoctoral Researcher. However, in exceptional cases, applicants who have completed almost all doctoral studies required for a PhD in a relevant field may also be considered. In such cases, the job title at the start of the employment will be Project Researcher and salary will be in the beginning on average 3100–3300 euros/month, depending on previous experience and skills. A person selected for the post of Project researcher is required to have a higher university degree.
We seek highly motivated, enthusiastic and hard-working candidates. The applicants must show good interpersonal skills and be willing to work in close collaboration with the project PI and other members of the international and multidisciplinary Human Diversity team, as well as have the ability to work independently.
We value equality and diversity in our work community and encourage qualified applicants, regardless of background, to apply for our open positions.
