Section 3 mentions "More specifically, this algorithm considers whether to assign a Stack Overflow comment j to a cluster C by examining all previously assigned comments Ci in that cluster and treating the classifier's prediction weight, wij >= 0:5, as a vote for j." However, Section 5 denotes this under the graph cutting algorithm "We use a lower voting algorithm threshold of wij >= 0:4 for a positive vote". Shouldn't this two be the same weight? Also, if you use 0.4 as the weight, I suggest that this threshold value should be justified using proper and concrete empirical evidence in Section 5. Otherwise, this looks like an arbitrary threshold.
Section 3 mentions "More specifically, this algorithm considers whether to assign a Stack Overflow comment j to a cluster C by examining all previously assigned comments Ci in that cluster and treating the classifier's prediction weight, wij >= 0:5, as a vote for j." However, Section 5 denotes this under the graph cutting algorithm "We use a lower voting algorithm threshold of wij >= 0:4 for a positive vote". Shouldn't this two be the same weight? Also, if you use 0.4 as the weight, I suggest that this threshold value should be justified using proper and concrete empirical evidence in Section 5. Otherwise, this looks like an arbitrary threshold.