Pseudo-Labels Are All You Need

Bogdan Kostić, Mathis Lucka, Julian Risch


Abstract
Automatically estimating the complexity of texts for readers has a variety of applications, such as recommending texts with an appropriate complexity level to language learners or supporting the evaluation of text simplification approaches. In this paper, we present our submission to the Text Complexity DE Challenge 2022, a regression task where the goal is to predict the complexity of a German sentence for German learners at level B. Our approach relies on more than 220,000 pseudolabels created from the German Wikipedia and other corpora to train Transformer-based models, and refrains from any feature engineering or any additional, labeled data. We find that the pseudo-label-based approach gives impressive results yet requires little to no adjustment to the specific task and therefore could be easily adapted to other domains and tasks.
Anthology ID:
2022.germeval-1.6
Volume:
Proceedings of the GermEval 2022 Workshop on Text Complexity Assessment of German Text
Month:
September
Year:
2022
Address:
Potsdam, Germany
Editors:
Sebastian Möller, Salar Mohtaj, Babak Naderi
Venue:
GermEval
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
33–38
Language:
URL:
https://aclanthology.org/2022.germeval-1.6
DOI:
Bibkey:
Cite (ACL):
Bogdan Kostić, Mathis Lucka, and Julian Risch. 2022. Pseudo-Labels Are All You Need. In Proceedings of the GermEval 2022 Workshop on Text Complexity Assessment of German Text, pages 33–38, Potsdam, Germany. Association for Computational Linguistics.
Cite (Informal):
Pseudo-Labels Are All You Need (Kostić et al., GermEval 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.germeval-1.6.pdf
Data
KlexikonTextComplexityDE