Abstract:
Product reviews can provide incredible advantages to customers and manufacturers.
Reviews may well be extending from a few to thousands that contains totally
different opinions. These build the method of analysing and extracting data
existing reviews become significantly and progressively difficult. In the first module
ot this project, sentiment analysis was accustomed analyse and extract sentiment
polarity on product reviews supported a particular feature of the product. This
analysis was conducted within the phases, like knowledge of pre-processing that
involves removal ofstopwords, and digits and punctuation, and part-of-speech (POS)
labelling, Term Frequency - Inverse Document Frequency (TF-IDF) and Naive
Bayes classifier is applied for the classification of sentiment polarity. In the second
module, the task was to detect and filter out the spam from a review. Spam reviews
can be written for both purposes i.e. to promote or demote any product. A normal
user cannot identify whether a written comment is genuine or spam. Therefore, we
have developed a model, which aims to filter out spam reviews by normalizing the
polarity of each review. Data used in this project are approx. 230 product reviews,
taken from Amazon.com.
