Product Quality Assessment Based on Online Reviews

Introduction

Online e-commerce platforms such as Amazon.com, JD.com and Taobao.com, often provide consumers venues to share their opinions and experience quality on purchasing products online and vote whether a review is useful or not. The affiliated review management system often rank reviews by the published time, the length of review text, the usefulness in the form of votes, or the star ratings of reviewers (Ghose and Ipeirotis, 2011; Huang et al., 2015).

Reviews often describe the product information, service quality on sales or logistics and the voice of consumers' experience quality. For example, the reviews on a dress might refer to the features such as name, price, color, size and style. Meanwhile, a review may describe different aspects of product quality or service experience quality, and different stakeholders might be interested in different product aspects and have different preferences accordingly. A consumer browses many reviews and then judges and weighs the views of the reviewers on product quality. The reviews attached to different product categories often refer to distinct features, and can embody the product quality to some extent. More and more consumers tend to read online product reviews to get helpful information about the quality of products or services before placing an order in e-commerce platforms (Sher and Lee, 2009), and the reviews significantly affect the consumers' decisions (Zhang et al., 2010; Zhang et al., 2014).

In e-commerce platforms, the benefit of reviews to decision makers depends heavily on review quality (Li et al., 2013; Ngo-Ye and Sinha, 2014). However, there are many unreliable and irrelevant reviews for malicious manipulations and frauds in websites (Hu, Liu and Sambamurthy, 2011; Liu et al., 2016), and the seemingly helpful information can often get from the reviews with more votes or the metadata in the form of features designated by e-commerce platforms. The quantity of available reviews on a product can be overwhelming and the quality of reviews tends to be very uneven, ranging from excellent detailed opinions to simple repetition of product specifications, in the worst case to pure spams. As the quantity of available reviews increases, it is almost impossible for a single end-user to browse all the reviews and take advantage of them. The abundant and uneven reviews may impede the stakeholders' ability to assess the quality of products or services and then to make effective decisions, so it is imperative and valuable to provide consumers helpful information on product quality directly.

However, the existing researches mainly focus on such as the assessment of review quality, the detection of fake reviews and spammers, the estimation of consumer preferences, the extraction of features, the mining of opinions, and there are few researches to assess product quality from the perspectives of product features based on reviews objectively and quantificationally. Therefore, the authors propose a method to assess product quality based on reviews in a granularity of product feature. Review quality assessment is the basis and premise of product quality assessment. The authors firstly define the related quality dimensions and develop the corresponding assessment models to assess the quality of reviews, and then extract product features and opinion words from the quality reviews, and finally assess the product quality on the extracted and consumer-concerned features.

The rest of this paper organized as follows. In section Introduction, the authors introduce the background of product quality assessment based on online reviews. In section Background, the authors analyze the product quality assessment based on reviews. The next section shows the architecture of product quality assessment, presents the extraction of features and opinion words, and develops the corresponding models for product quality assessment. In final section, the authors show and analyze the statistical and experimental results.