F. Niu, C. Zhang, C. Re & J. Shavlik (2012).
Elementary: Large-scale Knowledge-base Construction via Machine Learning and Statistical Inference. International Journal on Semantic Web and Information Systems (IJSWIS).

This publication is available in PDF.

Abstract:

Researchers have approached knowledge-base construction (KBC) with a wide range of data resources and techniques. We present Elementary, a prototype KBC system that is able to combine diverse resources and different KBC techniques via machine learning and statistical inference to construct knowledge bases. Using Elementary, we have implemented a solution to the TAC-KBP challenge with quality comparable to the state of the art, as well as an end-to-end online demonstration that automatically and continuously enriches Wikipedia with structured data by reading millions of webpages on a daily basis. We describe several challenges and our solutions in designing, implementing, and deploying Elementary. In particular, we first describe the conceptual framework and architecture of Elementary, and then discuss how we address scalability challenges to enable Web-scale deployment. First, to take advantage of diverse data resources and proven techniques, Elementary employs Markov logic, a succinct yet expressive language to specify probabilistic graphical models. Elementary accepts both domain-knowledge rules and classical machine-learning models such as conditional random fields, thereby integrating different data resources and KBC techniques in a principled manner. Second, to support large-scale KBC with terabytes of data and millions of entities, Elementary leverages high-throughput parallel computing infrastructure such as Hadoop, Condor, and parallel databases. Furthermore, to scale sophisticated statistical inference, Elementary employs a novel decomposition-based approach to Markov logic inference that solves routine subtasks such as classification and coreference with specialized algorithms. We empirically show that this decomposition-based inference approach achieves higher performance than prior inference approaches. To validate the ectiveness of Elementary's approach to KBC, we experimentally show that its ability to incorporate diverse signals has positive impacts on KBC quality.

Return to the publications of the Univ. of Wisconsin Machine Learning Research Group.