Speaker
Description
The aim of the paper is to evaluate automatic generation of specific kind of tests for English language learners. The kind of tests that we explore is referred to as cloze tests which are paragraphs of text with gaps which should be filled in by the learners. In the paper, we
explore efficiency of recurrent neural networks and transformer networks (BERT and ELECTRA) compared different quality metrics. The authors make the training and testing datasets available publicly. The general idea is based on paper by Felice at al.\cite{b1}. In the presented research, we extend the loss function with the Kullback–Leibler term and apply extra metrics.
The results present a promising quality of generated test for implementation in an online tool for the English language teachers and students.
