TY  - JOUR
T1  - Privacy Preserving Mining of Web Reviews Based on Sentiment Analysis and Fuzzy Sets
AU - A. Nofal, Mostafa AU - F. Sabbeh, Sahar AU - M. Fouad, Khaled 
JO  - Asian Journal of Information Technology
VL  - 19
IS  - 7
SP  - 122
EP  - 136
PY  - 2020
DA  - 2001/08/19
SN  - 1682-3915
DO  - ajit.2020.122.136
UR  - https://makhillpublications.co/view-article.php?doi=ajit.2020.122.136
KW  - Sentiment analysis
KW  -sentiments classification
KW  -privacy preserving
KW  -sentiment feature extraction
KW  -fuzzy sets
AB  - Customers like online submitting unstructured
reviews that has turned out a popular way to come across
sentiments across the products purchased or services
extradited. In each review, customer typically submit
about both the positive and negative opinion of the
product, although, the general sentiment toward that
product may be positive or negative. The sentiment
analysis tries to extract sentiments and subjectivity of
customer reviews. These reviews can be beneficial for
gathering sentiments of customers about products by
analyzing it. However, this analysis should derive careful
consideration of customer’s anonymity and the privacy of
the sensitive data because a privacy is a significant
concern for either customers and enterprises. In this
research, automatic analysis of sentiment is carried out to
achieve such detailed aspects based on domain ontology.
Sentiment analysis recognizes the features in the
sentiment and classify the sentiments of the review for
each of these features. In the proposed approach, the
sentiment polarity and polarity strength are provided and
computed using fuzzy set. The fuzzy set theory is just
effective in processing natural languages because it
measures the vagueness. The fuzzy set theory is effective
in analyzing reviews which are generally in natural
languages. Additionally, the proposed system takes
privacy into consideration by masking data before final
publishing. The evaluation of the proposed approach is
based on using dataset of London restaurant’s reviews on
TripAdvisor. The evaluation utilizes three different
classifiers MLP, SVM and NB and utilizes 5×2-fold cross
validation for four evaluation measures; accuracy,
precision, recall and F1.
ER  -