PolQA: Polish Question Answering Dataset
This addresses the lack of QA datasets for Polish, enabling research and applications in a new language domain, though it is incremental as it adapts existing methods to a new dataset.
The authors introduced PolQA, the first Polish dataset for open-domain question answering, consisting of 7,000 questions and over 7 million candidate passages, and proposed an annotation strategy that increased passage retrieval accuracy@10 by 10.55 percentage points while reducing annotation cost by 82%.
Recently proposed systems for open-domain question answering (OpenQA) require large amounts of training data to achieve state-of-the-art performance. However, data annotation is known to be time-consuming and therefore expensive to acquire. As a result, the appropriate datasets are available only for a handful of languages (mainly English and Chinese). In this work, we introduce and publicly release PolQA, the first Polish dataset for OpenQA. It consists of 7,000 questions, 87,525 manually labeled evidence passages, and a corpus of over 7,097,322 candidate passages. Each question is classified according to its formulation, type, as well as entity type of the answer. This resource allows us to evaluate the impact of different annotation choices on the performance of the QA system and propose an efficient annotation strategy that increases the passage retrieval accuracy@10 by 10.55 p.p. while reducing the annotation cost by 82%.