News: LREC-COLING 2024 Presentation
#post
Happy to share that I will be presenting the paper “ Improved Neural Protoform Reconstruction via Reflex Prediction” next month at LREC-COLING 2024 in Italy.
Since the 19th century, historical linguists have developed the comparative method, enabling the reconstruction of ancestors of families of languages (called proto-languages) by comparing related words in descendent languages (called reflexes).
Manual comparative reconstruction is a laborious process, so computational linguists have attempted to automate protolanguage reconstruction with neural networks.
Existing models, however, mostly treat reconstruction as a generic sequence-to-sequence task, without incorporating the comparative method’s constraint that descendent languages can be derived from the reconstruction.
We propose a reconstruction workflow in which candidate reconstructions are reranked by a reflex prediction model. We show that this more complete implementation of the comparative method allows us to achieve better reconstruction performance.
It has been a great time working with Helen Wang and David Mortensen on this very challenging task in NLP!
To find out more: