This data set accompanies the research paper, Automating Behavioral Testing in Machine Translation.
It includes the following:
-
data/prompts_src_gen
contains the prompts that were used to generate the English source sentences for behavioral testing, as described in the paper. -
data/filtered_src_gen
contains the generated source sentences after the filter steps outlined in the paper. -
data/prompts_candidates
contains the prompts that were used to generate the target-language-specific candidate sets. -
data/candidate_sets
contains the generated candidate sets (unfiltered).
If you use this dataset, please cite our paper as follows:
Javier Ferrando, Matthias Sperber, Hendra Setiawan, Dominic Telaar, Saša Hasan (2023). Automating Behavioral Testing in Machine Translation. Conference on Machine Translation (WMT).