Using Markov Logic to Refine an Automatically Extracted Knowledge Base

A number of information extraction (IE) projects such as NELL and TextRunner seek to build a usable knowledge base from the rapidly growing amount of information on the web. However, these solutions use heuristic approaches to reasoning rather than sound probabilistic inference. In this paper, we present a method based on Markov logic for cleaning an automatically extracted knowledge base using only the confidence values and ontological constraints of the original system. Our approach works by reasoning jointly over all candidate facts. To achieve scalability, we introduce a neighborhood grounding method that only instantiates the part of the network most relevant to the given query. This allows us to partition the knowledge cleaning task into tractable pieces that can be solved individually. In experiments on NELL’s knowledge base, our method improves both F1 and AUC.