Module Extraction from Uncertain Knowledge Graphs

Supervisor: Ameneh Naghdi Pour (a.naghdipour@vu.nl, j.y.chen@vu.nl)

Project Description

Knowledge graphs have been increasingly applied in many domains, including manufacturing and production systems, robotics and autonomous systems, healthcare and biomedical engineering and many others. One of the main benefits and reasons for exploiting knowledge graph is that they enable efficient information retrieval, allowing users to extract a targeted subset of the knowledge graph that satisfies their information needs. In many real-world and industrial applications, however, domain knowledge is inherently uncertain. Assigning weights to facts makes it possible to express their relative confidence, likelihood, or evidential strength, enabling reasoning and diagnosis methods to rank competing hypotheses and prioritise the conclusions best supported by the available knowledge. The goal of this project is to extract the most plausible subset of such uncertain knowledge graphs, given a set of entities of interest.

Task

This project will involve researching existing module extraction methods in RDF knowledge graph, developing an algorithm that extract module from an uncertain knowledge graph, given multiple entities of interest, and finally evaluating the proposed algorithm on an aircraft knowledge graph.

Research question

What is the most plausible module in an uncertain RDF knowledge graph given multiple entities of interest?