Skip to main content

Table 11 Performance comparison with other methods

From: SentiHealth: creating health-related sentiment lexicon using hybrid approach

Work

Dataset

Noise Reduction

Features

Approach

Precision

Recall

F-measure

Goeuriot et al. (2012)

25,000 reviews

Filtering

Unigram and bigram

Hybrid (lexicon-based + information gain)

0.76

0.52

0.62

Asghar et al. (2015)

15,000 reviews

Filtering, tokenization, stop word removal, stemming

Unigram, bigram and trigram

Supervised (revised mutual information)

0.78

0.64

0.64

Demiroz et al.(2012)

9000 reviews

Filtering, stop word removal

Bag of words

Supervised (delta scoring)

0.75

0.48

0.58

Our work

26,060 reviews

Filtering, tokenization, stop word removal, lemmatization, spell correction, co-reference resolution

Unigram, bigram and trigram

Hybrid (boot strapping + corpus-based)

0.89

0.79

0.83

  1. All the mentioned evaluations measures are as reported by their respective work