Etymon Project

Research Focus:

We develop computational methods for modeling

• language evolution, and
• relationships among languages within language families.

The methods are based on the information-theoretic Minimum Description Length Principle (MDL). The methods were initially tested on the Uralic language family—viz., the languages genetically related to Finnish. The methods are applicable generally; we currently explore their application to the following language families:

• Uralic
• Turkic
• Indo-European
• Khoisan

Supported by:

• Academy of Finland, the UraLink Project,
• Russian Fund for the Humanities / Russian Foundation for Basic Research,
• HIIT: Helsinki Institute for Information Technology,
• Algodan Center of Excellence: Algorithmic Data Analysis.

Language evolution and etymology:

• Etymon overview (poster)
• Etymon Modeling Suite
  • Documentation

Unsupervised Learning of Morphology:

• Evaluation package from the LREC-2016 paper

Transliteration:

• Cross-lingual Transliteration Dataset

Resources:

• StarLing: collection of etymological databases
  for many language families of the world.
• Web-based etymological database,
  Suomen Sanojen Alkuperä (SSA):
  Finnish Etymological Dictionary
  (To be published on the Web by Kotus,
  please contact us to request access)
• Wiki: Etymon Project (Internal use)
• Wiki: Computational historical linguistics
  and modeling population history (Internal use)

People:

• Mian Du, PhD student
• Suvi Hiltunen, MSc student
• Guowei Lv, MSc student
• Javad Nouri, MSc student
• Kirill Reshetnikov,
  Russian Academy of Sciences,
  Institute of Linguistics, Moscow
• Arto Vihavainen, MSc student
• Marjaana Välisalo, MSc student
• Hannes Wettig, PhD student
• Roman Yangarber: Project Lead

Collaboration:

• Russian Academy of Sciences (RAS), Institute of Linguistics.
• Bayesian Statistics Group, led by J Corander, Department of Mathematics and Statistics, COIN Center of Excellence of the Academy of Finland. Applying models from population-genetics to linguistic data to model language evolution.
• KOTUS: analysis and enhancement of data in the Finnish etymological dictionary "Suomen Sanojen Alkuperä". (This database is proprietary, and will be released for public access soon. Please contact us or KOTUS to request permission to access.)

Publications: conference and journal papers, book chapters, dissertations

1. Modeling language evolution with codes that utilize context and phonetic features   (pdf)
   
   Javad Nouri, Roman Yangarber
   In Proceedings of CoNLL: 2016 Conference on Computational Natural Language Learning
   (2016) Berlin, Germany

2. From alignment of etymological data to phylogenetic inference via population genetics   (pdf)
   
   Javad Nouri, Jukka Sirén, Jukka Corander, Roman Yangarber
   In Proceedings of CogACLL: the 7th Workshop on Cognitive aspects of Computational Language Learning colocated with ACL-2016
   (2016) Berlin, Germany

3. Minimum Description Length Models for Unsupervised Learning of Morphology   (Master's Thesis)
   
   Javad Nouri
   (2016) University of Helsinki, Department of Computer Science

4. A novel method for evaluation of morphological segmentation   (pdf)
   
   Javad Nouri, Roman Yangarber
   In Proceedings of LREC: 10th International Conference on Language Resources and Evaluation
   (2016) Portorož, Slovenia

5. Measuring Language Closeness by Modeling Regularity   (pdf)
   
   Javad Nouri, Roman Yangarber
   In Proceedings of the EMNLP 2014 Workshop on Language Technology for Closely Related Languages and Language Variants
   (2014) Doha, Qatar

6. Cognate discovery and alignment in computational etymology   (Master's Thesis)
   
   Guowei Lv
   (2014) University of Helsinki, Department of Computer Science

7. MDL-based Models for Transliteration Generation   (pdf)
   
   Javad Nouri, Lidia Pivovarova, Roman Yangarber
   SLSP 2013: International Conference on Statistical Language and Speech Processing
   Springer Verlag, Lecture Notes in Artificial Intelligence (LNAI) Volume 7978, (2013) Tarragona, Spain

8. Information-theoretic modeling of etymological sound change   (abstract)
   
   Hannes Wettig, Javad Nouri, Kirill Reshetnikov and Roman Yangarber
   Invited chapter in Approaches to measuring linguistic differences (Lars Borin, Anju Saxena, eds.) Trends in Linguistics Series, Volume 265.
   (2013) Mouton de Gruyter

9. Probabilistic, Information-Theoretic Models for Etymological Alignment   (Ph.D. Thesis)
   Hannes Wettig
   (2013) University of Helsinki, Department of Computer Science

10. Information-theoretic Methods for Analysis and Inference in Etymology   (pdf)
    
    Hannes Wettig, Javad Nouri, Kirill Reshetnikov and Roman Yangarber
    In Proceedings of WITMSE-2012: the 5th Workshop on Information-theoretic Methods in Science and Engineering   (Steven de Rooij, Wojciech Kotłowski, Jorma Rissanen, Petri Myllymäki, Teemu Roos & Kenji Yamanishi, eds.)
    (2012) Amsterdam, the Netherlands

11. Minimum Description Length Modeling of Etymological Data   (Master's Thesis)
    
    Suvi Hiltunen
    (2012) University of Helsinki, Department of Computer Science

12. Using Context and Phonetic Features in Models of Etymological Sound Change   (pdf)
    
    Hannes Wettig, Kirill Reshetnikov and Roman Yangarber.
    In Conference of the European Chapter of the Association for Computational Linguistics (EACL) Workshop on Visualization of Linguistic Patterns and Uncovering Language History from Multilingual Resources
    (2012) Avignon, France

13. MDL-based models for alignment of etymological data   (pdf)
    
    Hannes Wettig, Suvi Hiltunen, Roman Yangarber.
    RANLP-2011: Conference on Recent Advances in Natural Language Processing
    (2011) Hissar, Bulgaria

14. MDL-based modeling of etymological sound change in the Uralic language family   
    
    Hannes Wettig, Suvi Hiltunen, Roman Yangarber.
    WITMSE-2011: The 4th Workshop on Information Theoretic Methods in Science and Engineering
    (2011) Helsinki, Finland

15. Probabilistic models for alignment of etymological data   (pdf)
    
    Hannes Wettig, Roman Yangarber.
    Nodalida-2011: Nordic Conference on Computational Linguistics
    (2011) Riga, Latvia

16. Hidden Markov models for induction of morphological structure of natural language   
    
    Hannes Wettig, Suvi Hiltunen, Roman Yangarber.
    WITMSE-2010: Workshop on Information Theoretic Methods in Science and Engineering
    (2010) Tampere, Finland

Sample reconstruction of the Turkic Family, based on StarLing data: