Datasets:
Updated extended description
Browse filesThe Combinations of Aligned data-sentenCes from nAturally PrOduced texts (hereafter: CACAPO) dataset is a dataset for data-to-text generation, which contains automatically scraped news reports for the sports, weather, stock, and incidents domain in English and Dutch, aligned with relevant attribute-value paired data, which were reverse-engineered from the articles. This dataset contains the original articles and attribute-value pairs for end-to-end training, where the training split has been manually altered to elongate 1504 (10% of the dataset) input records with contextually relevant attribute value pairings.