Doremi Research Group: Computational Linguistics, Language Technology, Digital Humanities and Social Sciences

The Group conducts research on problems of analyzing linguistic data in various application domains:

• how language conveys information,
• how information can be extracted from text,
• how hidden, underlying structure can be learned from observed data.

Our work combines empirical and theoretical approaches to these problems, through research projects with international and domestic partners in academia, industry and government organizations.

The Research Group is part of the Centre of Excellence ALGODAN: Algorithmic Data Analysis of the Academy of Finland.
Group leader: Dr Roman Yangarber.

• PULS:
• Research: Semantic analysis of text for surveillance of news media. Information Extraction (IE) is a language-understanding task, whose objective is to find facts or events in text, collect them into a data base, and perform reasoning over the collected data. Methods: unsupervised and weakly-supervised learning, neural nets. Domains covered include:
  • business intelligence,
  • epidemiological surveillance,
  • cross-border security and criminal activity.
• Collaboration: Please see project page.
• Funding:
  • Tekes
  • European Commission


• Revita:
  • Research: computational methods for supporting language learning, and for supporting revitalization for endangered languages. Initial focus on Finno-Ugric languages, but now covering several others.
  • Collaboration:
    • Yle,
    • Opetushallitus.
  • Funding: funded by the Academy of Finland through project FinUgRevita.
  
  
• Etymon:
  • Research: computational models of language evolution, modeling how the Finnish language genetically relates to the Uralic language family, based on lexical data in etymological databases. Methods: information-theoretic principles, including the Minimum Description Length principle (MDL). Applying the methods more generally, beyond the Uralic family—Turkic, Indo-European and Khoisan.
  • Collaboration:
    • The Russian Academy of Sciences (RAS), Institute of Linguistics.
    • The StarLing Project, a collection of etymological databases for many language families of the world.
    • KOTUS: enhancement of the Finnish etymological dictionary "Suomen Sanojen Alkuperä". (The database is proprietary, and will be released for public access soon.)
  • Funding: jointly funded by the Academy of Finland and the Russian Fund for the Humanities, through project UraLink.
  
  
• Etymological BANANAS:
  • Research Focus: Modeling genetic relationships among members of a language family, using methods from population genetics. Application to etymological data from different language families, starting with Uralic and Turkic.
  • Collaboration:
    • J Corander's group, at the Department of Mathematics and Statistics: population-genetics models. Part of the COIN Center of Excellence of the Academy of Finland
    • Russian Academy of Sciences: analysis of etymological databases, the StarLing Project
    • KOTUS: etymological databases of the Uralic language family.
  • Funding: A HIIT Pump-Priming Project

  ---

  Prior projects

• Clarin:
  • the EU Clarin project for building infrastructures for linguistic resources.
• ContentFactory:
  • Collaboration between PULS and the TermFactory project at the Department of Modern Languages, focusing on large-scale ontology creation and maintenance, for text-analysis tasks.