Refining Automatically Extracted Knowledge Bases Using Crowdsourcing

Algorithm 1

Rank-based knowledge refining.

Input: A set of extracted candidate facts , semantic constraints , a budge , a threshold

Output: Refined knowledge base

Initialize with machine based estimations

Calculate scores using Eq. (15)

Calculate scores using Eq. (17)

Calculate using Eq. (18)

Rank candidate facts by scores and select top instances from to conduct crowdsourcing

facts with a confirm from the crowd