IBM’s Type Prediction Systems Eliminate Need for Manual Annotations on Knowledge Graphs

Knowledge graphs (KGs) are graphs used to accumulate and convey real-world knowledge. KG nodes capture information about entities of interest (like people, places or events) in a given domain or task, while the edges represent the connections between them. To provide vital information for related tasks such as Knowledge Base Question Answering (KBQA), various semantic web technologies have been employed to represent KGs with explicit semantics, defining a type for each node. A “Tylor Swift” node for example could be classified as a “popular singer” type. Although predicting KG type information is crucial for solving KG-related and downstream tasks, most existing work in this area uses supervised solutions that operate on relatively small-to-medium-sized systems.

In the paper Type Prediction Systems, the IBM researchers introduce two systems for predicting type information at any granularity and without annotations. Their TypeSuggest module is an unsupervised system designed to generate types for a set of seed query terms input by the user, while the Answer Type prediction module predicts the correct answer type for user-provided questions.