SailGenie Dataset

The dataset contains suitable data in the scenario of generating a Knowledge Graph in the world of sailing.

The dataset includes the original corpus, i.e., the set of textual excerpts extracted from several information sources, a suitable ground truth, i.e., a set composed of domain triplets manually inferred by the corpus, and a set of triplets generated with the OIE-based tools, annotated and evaluated by human assessors.

The two folders identify the two sailing sub-domains:
 - `BASICS` related to fundamental knowledge for beginners about navigation, behavior, and maneuvers;
 - `SAFETY`: includes information on measures, legal requirements, instruments, and best practices specifically required to ensure everyone's safety during sailing.

Each folder has the following data:
 - `sentences.txt` contains the collected sailing excerpts in the form o an ordered list of sentences;
 - `ground_truth_triplets.csv` contains the manually identified triplets from the collected sentences;
 - `annotaded_triplets.csv` contains the manually supervised sample of automatically extracted triplets, all labeled as valid or invalid.

The `sen_index` field in both `ground_truth_triplets.csv` and `annotated_triplets.csv` refers to a line number in `sentences.txt` (start counting from zero) where to find the sentence related to each triplet.