The team of Anastasia Bystrova, an employee of ANR-Lab, spoke at the Dialogue-2023 conference

The researchers made a presentation in the main program of Dialogue 2023, the leading conference on computational linguistics in Russia. Their article Parameter-Efficient Tuning of Transformer Models for Anglicism Detection and Substitution in Russian was published in the conference proceedings.

Daniil Lukichev, Daria Kryanina, Anastasia Bystrova, Alena Fenogenova and Maria Tikhonova created a dataset of >1000 pairs of parallel sentences: one of the sentences in the pair is with actual Anglicism, the second one is with the Russian counterpart. This dataset has been downloaded more than 100 times. 

Using this dataset, two models were trained:

• model for detecting anglicisms
• model for paraphrasing anglicisms 

These models are especially relevant in the light of the recent change in the Federal Law 'On the State Language of the Russian Federation', which restricts the use of anglicisms.

In their research, Anastasia and her colleagues relied on publications of ANR-Lab researcher Ilia Karpov: Comparative Analysis of Anglicism Distribution in Russian Social Network Texts and A General Method Applicable to the Search for Anglicisms in Russian Social Network Texts.