Constructing Bayesian networks by harvesting knowledge from online resources

In this paper, the development of a human-like intelligent system, named AKEOS (Automatic Knowledge Extraction from Online Sources), is introduced. AKEOS can automatically harvest knowledge from online resources to build a Bayesian network inference engine. Starting from a single event, the AKEOS system performs unsupervised knowledge extraction to convert unstructured text into structured knowledge. By performing repeated knowledge extraction for multiple similar events, AKEOS produces structured databases. Thus, various machine learning algorithms can be directly applied to explore relations between attributes, to discover patterns hidden in the data, or to build inference engines such as Bayesian networks for predictive and diagnostic reasoning. AKEOS is an end-to-end system with lists of events as input, structured databases as intermediate products, and inference engines as end products.