Refining Automatically Extracted Knowledge Bases Using Crowdsourcing

Algorithm 6

Error-tolerant graph coloring.

Input:

Output: All vertices in are colored as Green or Red

  while  there exist uncolored vertices in   do

Select a set of uncolored vertices to conduct crowdsourcing;

for  each crowdsourced with an answer  do

if    then

color and related vertices using coloring strategy;

else

color Blue;

end if

end for

  end while

Learn a logistic regression model to the machine confidence scores from Green and Red vertices;

 Predict colors of facts in Blue vertices;

 return colored