Task C - Non-Taxonomic Relationship Extraction

Identify non-taxonomic, semantic relations between types.

Training instances will be given as

with Th and Tt being head and tail taxonomic types, respectively, and r the non-taxonomic semantic relation between them, chosen from a predefined set R of semantic relations. Participants will train a system to classify pairs of types into these semantic relations. The training phase involves types and triples of semantic relations; the test phase requires applying the trained system to predict semantically related triples from given types and the set of relations. The caveat here is that we do not expect participant systems to infer a semantic relation but rather establish semantically related types and classify their relation from a known set of predetermined relations.

SubTask C.1 - Non-Taxonomic Relationship Extraction - UMLS

Non-Toxonomic Relationship Extraction - UMLS subtask facilitates the extraction of non-taxonomic relationships, which involves identifying and establishing connections between medical concepts that are not directly hierarchical in nature. These relationships are essential for capturing complex interactions and associations between medical entities, such as diseases, symptoms, treatments, and anatomical structures. Statistics of train, and test set within this task are described as follows:

SubTask C.2 - Non-Taxonomic Relationship Extraction - GO

Non-Toxonomic Relationship Extraction - Gene Ontology subtask facilitates the formal representation of a body of knowledge within biological domain. The structure of GO can be described in terms of a graph, where each GO term is a node, and the relationships between the terms are edges between the nodes.