User Input-Based Construction of Personal Knowledge Graphs

Personal knowledge plays a key role in the development of more intelligent applications. Applying knowledge representation techniques like knowledge graphs to the representation of personal knowledge is under active research. However, current knowledge graph construction methods are hindered by problems like absence of knowledge, ambiguity, conflicts and erroneous knowledge when applied to personal knowledge. This is largely due to its unique properties, such as its user-specific, volatile nature and limited data availability. We present in this paper a novel method supporting user input-based construction of personal knowledge graphs. We develop a new knowledge graph structure specifically to counter the said problems, and present a method that uses an iteration-specific subgraph as the intermediate layer between the user and the actual personal knowledge graph for better integration of user input. We also propose a deprecation mechanism to address the volatile nature of personal knowledge.