Low-Resource Grammatical Error Correction

Organisatsiooni nimi
TartuNLP
Kokkuvõte
The best performing grammatical error correction methods are English centred and rely on larger error correction corpora than is available for Estonian and many other languages. For Estonian, we have corrections for less than 10k erroneous sentences. The aim is to research methods that can use additional resources, like translation examples or text with synthetic errors, for better error correction. Proposing your own idea is encouraged, but we can also find a more specific topic together, considering previous knowledge and interests.

See more about the work done in TartuNLP from https://arxiv.org/abs/1903.11283 or find previous bachelor and master theses on grammatical error correction or contact for more information.
Lõputöö kaitsmise aasta
2022-2023
Juhendaja
Agnes Luhtaru
Suhtlemiskeel(ed)
eesti keel, inglise keel
Nõuded kandideerijale
Knowing basic Python, Linux and machine learning is essential, but the difficulty depends on the specific topic.
Tase
Bakalaureus, Magister
Märksõnad

Kandideerimise kontakt

 
Nimi
Agnes Luhtaru
Tel
E-mail
agnes.luhtaru@ut.ee