Table 2A

Studies that performed sentiment analysis using supervised approach, including the number of raters and associated inter-rater agreement expressed as Cohen’s kappa (κ), classifiers and configuration applied where reported. Studies are reported in chronological order

AuthorData sourceComments classifiedNo. of ratersκSentiment categoriesClassifierConfiguration
Alemi et al34RateMDs100%*
NRNRSparsity rule
Information gain
SVM: RBF kernel
Greaves et al7NHS choices17.56%† (1000/5695)20.76Prior polarity
Information gain
SVM: RBF kernel
Wagland et al48Cancer experience14.19% (800/5634)30.64–0.87NR
Bahja et al26NHS choices75% (56 818/76 151)N/AN/ASparsity rule
Ratings in binary sentiment
Jimenez-Zafra et al54COPOS and COPOD‡100% (n=156 975 COPOD and n=743 COPOS)N/AN/ARatings in binary sentiment
SVM: linear kernel
Huppertz et al6Facebook0.88% (508/57 986)3NRNR
Doing-Harris et al24Press Ganey0.58% (300/51 235)30.73NR
Menendez et al47Vendor supplied100% (132/132)NRNRNR
  • *Classified as praise (positive), complaint (negative), praise and complaint (mixed), neither (neutral).

  • †Only n-grams classified.

  • ‡Also used dictionary lookup and cross domain method.

  • B, bagging; COPOD, corpus of patient opinions in Dutch; COPOS, corpus of patient opinions in Spanish; DT, decision trees; GL, generalised linear model; KN, k-nearest neighbour; NB, Naïve Bayes; NR, not reported; RBF, radial basis function; RF, random forest; SVM, support vector machine.