no code implementations • 26 May 2023 • Vít Novotný, Kristýna Luger, Michal Štefánik, Tereza Vrabcová, Aleš Horák
Although pre-trained named entity recognition (NER) models are highly accurate on modern corpora, they underperform on historical texts due to differences in language OCR errors.
1 code implementation • 24 May 2023 • Marek Kadlčík, Michal Štefánik, Ondřej Sotolář, Vlastimil Martinek
We address this deficiency by creating Calc-X, a collection of datasets that demonstrates the appropriate use of a calculator in reasoning chains.
no code implementations • 23 May 2023 • Michal Štefánik, Marek Kadlčík
Many recent language models (LMs) of Transformers family exhibit so-called in-context learning (ICL) ability, manifested in the LMs' ability to modulate their function by a task described in a natural language input.
no code implementations • 11 May 2023 • Lukáš Mikula, Michal Štefánik, Marek Petrovič, Petr Sojka
We find that while existing debiasing methods can mitigate reliance on a chosen spurious feature, the OOD performance gains of these methods can not be explained by mitigated reliance on biased features, suggesting that biases are shared among different QA datasets.
1 code implementation • 4 Apr 2023 • Michal Štefánik, Marek Kadlčík, Piotr Gramacki, Petr Sojka
Despite the rapid recent progress in creating accurate and compact in-context learners, most recent work focuses on in-context learning (ICL) for tasks in English.
no code implementations • 3 Dec 2022 • Michal Štefánik, Marek Kadlčík
We find that most of the recent in-context learners can not consistently benefit from the demonstrated concepts, irrespective of the model size.
1 code implementation • 29 Nov 2022 • Michal Štefánik, Marek Kadlčík, Petr Sojka
Domain adaptation allows generative language models to address specific flaws caused by the domain shift of their application.
no code implementations • 16 Jun 2022 • Michal Štefánik
Despite their outstanding performance, large language models (LLMs) suffer notorious flaws related to their preference for simple, surface-level textual relations over full semantic complexity of the problem.
1 code implementation • ACL 2022 • Michal Štefánik, Vít Novotný, Nikola Groverová, Petr Sojka
Progress in natural language processing research is catalyzed by the possibilities given by the widespread software frameworks.
1 code implementation • WMT (EMNLP) 2021 • Michal Štefánik, Vít Novotný, Petr Sojka
This work introduces a simple regressive ensemble for evaluating machine translation quality based on a set of novel and established metrics.
no code implementations • 1 Jun 2021 • Dávid Lupták, Vít Novotný, Michal Štefánik, Petr Sojka
Math informational retrieval (MIR) search engines are absent in the wide-spread production use, even though documents in the STEM fields contain many mathematical formulae, which are sometimes more important than text for understanding.
1 code implementation • 19 Apr 2021 • Vít Novotný, Michal Štefánik, Eniafe Festus Ayetiran, Petr Sojka, Radim Řehůřek
In 2018, Mikolov et al. introduced the positional language model, which has characteristics of attention-based neural machine translation models and which achieved state-of-the-art performance on the intrinsic word analogy task.
no code implementations • RANLP 2021 • Vít Novotný, Eniafe Festus Ayetiran, Dalibor Bačovský, Dávid Lupták, Michal Štefánik, Petr Sojka
In our work, we find the optimal subword sizes on the English, German, Czech, Italian, Spanish, French, Hindi, Turkish, and Russian word analogy tasks.
1 code implementation • 10 Mar 2020 • Vít Novotný, Eniafe Festus Ayetiran, Michal Štefánik, Petr Sojka
In our work, we investigate the individual and joint effect of the two word embedding regularization techniques on the document processing speed and the task performance of the SCM and the WMD on text classification.
Ranked #2 on Document Classification on Amazon