TurkuNLP is a multidisciplinary research group that combines digital linguistics, natural language processing (NLP), and machine learning. We specialize in developing state-of-the-art NLP tools for Finnish and other languages, as well as applying these methods to analyze how language is used in large-scale digital language data. For instance, we have developed the FinBERT and PORO language models and the Finnish Internet Parsebank dataset composed of billions of words.

We welcome applications for a fixed-term position / positions of a Postdoctoral Researcher / Project Researcher. The positions are for the period of December 1, 2024, to December 31, 2025, and can be either part-time or full-time.

The positions are part of two research projects. "Mechanisms of register variation in massively multilingual web-scale corpora” (Mecha) is funded by the Research Council of Finland. The project aims to explore language use in the digital world at a massively multilingual scale. The specific focus of the project is on web registers, such as news, blogs, and how-to pages. The primary goals are to analyze the linguistic characteristics of web registers across languages and to understand how and why both the set of registers we use on the web and the register characteristics they feature vary – not only across languages but also across cultures and countries. What are the mechanisms behind this variation? To achieve this, the project applies and develops forward state-of-the-art machine learning methods together with giant, multilingual web-crawled datasets. Ultimately, the project also applies the developed methods to provide metadata for web datasets and thus contribute to their usage in an informed manner. For more information, please visit turkunlp.org and refer to our latest publications.

“Fin-Clariah” is a national research infrastructure for Social Sciences and Humanities with partners in most Finnish universities. Our part in TurkuNLP focuses on modeling noisy web data and especially texts originating from social media. The applied data and methods are very similar to those in Mecha – our goal is to benefit from the results in Mecha and our previous projects to create better research infrastructure for all fields working on web-based language use.

Job description

The selected candidate will work as part of the Mecha and / or Fin-Clariah project teams and TurkuNLP. The candidate is expected to utilize and integrate diverse register datasets that have already been developed in the project, both manually and automatically labeled, and potentially develop new ones. The specific approaches and research questions will depend on the candidate's profile. Possible research questions include: What types of register characteristics are shared across languages, and which ones are unique to specific languages? How does the set of web registers and their characteristics depend on factors such as language, country or culture? Additionally, the candidate is expected to participate in the supervision of junior researchers and planning of funding proposals.