Fuzzy Aspect Based Opinion Classification System for Mining Tourist Reviews

Table 1

Critical evaluation of aspects extraction methods.

Reference	Explicit aspects		Implicit aspects	Coreferential aspects	Irrelevant aspects	Method	Aspects selection	Results
	Frequent	Infrequent
Marrese-Taylor et al., 2014 [19]	High	Null	Null	Not handled	Handled	Rules based	Frequent nouns	30%
Marrese-Taylor et al., 2013 [18]	High	Null	Null	Not handled	Handled	Rules based	Frequent nouns	Not given
de Albornoz et al., 2011 [17]	High	Null	Null	Handled	Handled	Rules based	Relative importance	66.8%
Muangon et al., 2014 [16]	High	Null	Null	Not handled	Handled	Rules based	Ranking	Not given
Pekar and Ou, 2008 [15]	High	Null	Null	Not handled	Handled	Rules based	Frequent nouns	Not given
Hai et al., 2014 [20]	High	Low	Null	Not handled	Not handled	Rules based	Domain-specific nouns	65%
Colhon et al., 2014 [21]	High	Low	Null	Not handled	Handled	Seeds based	Grammatical relationship	Not given
Mukherjee and Liu, 2012 [22]	High	Low	Null	Not handled	Handled	Seeds based	Higher-order cooccurrences	77%
Wang et al., 2010 [23]	High	Low	Null	Not handled	Handled	Seeds based	Maximum term overlapping	Not given
Zhu et al., 2011 [24]	High	Low	Null	Not handled	Handled	Seeds words	Frequency of cooccurrence	69%
Wu and Ester, 2015 [25]	High	Medium	Null	Not handled	Not handled	Topic model based	Connected topics	Not given
Xianghua et al., 2013 [26]	High	Medium	Null	Not handled	Not handled	Topic model based	Minimum distance with topics	73%
Xueke et al., 2013 [27]	High	Medium	Null	Not handled	Not handled	Topic model based	Frequent topics	Not given
Proposed method	High	Medium	High	Handled	Handled	Fuzzy based	FURIA rules	81%

We represent aspects into three types: frequent explicit aspects, infrequent explicit aspects, and implicit aspects shown in 2 to 4 columns of this table. Coreferential and irrelevant aspects handing shown in column 5 and column 6, respectively. We labeled these columns by null, low, medium, high, handled, and not handled.
“Null” = not extracting those types of aspects.
“Low” = extracting 10 to 40% of those types of aspects.
“Medium” = extracting 40 to 70% of those types of aspects.
“High” = extracting 70 to 100% of those types of aspects.
“Handled” = handling those types of aspects.
“Not handled” = not handling those types of aspects.