Two representative methods were used as baselines: the generative model proposed by (Brill and Moore, 2000) referred to as generative and the logistic regression model proposed by (Okazaki et al., 2008)

When using their method for ranking, we used outputs of the logistic regression model as rank scores.

(2008) proposed using a logistic regression model for approximate dictionary matching.

(2008) utilized substring substitution rules and incorporated the rules into a L1-regularized logistic regression model.

Both the string similarity classifier and the relation classifier are trained using Lg-regularized logistic regression .

As we trained both classifiers using logistic regression , we have models for the probabilities P(Y|X1) and P(Y|X2).

(typically poorly calibrated) probability estimates of logistic regression .

We ran all of our experiments in Weka (Hall et al., 2009) using logistic regression over 10 runs of 10—fold cross-validation.

We use logistic regression as our classifier because it has been shown that logistic regression typically has lower asymptotic error than naive Bayes for multiple classification tasks as well as for text classification (Ng and Jordan, 2002).

We experimented with an SVM classifier and found logistic regression to do slightly better.