| dc.description.abstract | In today`s era of easy access to information, online consumers have become more
informed in their decision making about the products they would like to buy. Online
product reviews have played a key role in the increase in consumer awareness and
online research activities about products. Due to the vast number of product reviews
(in thousands) for each item, it becomes cumbersome to makes sense of all the information and form a perspective or develop a sentiment about the product. In order
to tackle this problem, large websites such as Amazon provide a helpfulness score
along with each review, to help uninformed consumers get an idea of the authenticity, quality and perspective of a particular review, which are written by consumers
themselves having experience in purchasing or using that product.
We aim to study reviews from the Amazon product review dataset and under-
stand how various review attributes influence the review helpfulness score as well as,
how this influence varies across diverse product categories. For this purpose, we will
look at key statistical features from the star-ratings as well as context based features
extracted from the reviews. As an addition to our existing task, we will also discuss
possible origins of biases in the system and look at model building approaches that
can reduce the effect of intrinsic biases in a particular product's review helpfulness
voting activity. This research will contribute significantly towards understanding the
characteristics of helpful product reviews across different categories and lay foundations for future methods for preventing biased helpfulness voting on online product
review platforms. | en |
