Pavel Braslavsky, Senior researcher at Combinatorial Algebra Lab of Ural Federal University:
"High-quality annotated data is crucial for measurable progress in NLP tasks. In the field of knowledge base question answering there are no reliable automatic annotation tools to date. At the same time, in-house data annotation makes the dataset creation process slow and expensive. Thus, we used Toloka as the core of our data annotation pipeline and were able to create high-quality datasets for Russian knowledge base and hybrid QA evaluation."