Your reviews or mine? Exploring the determinants of “perceived helpfulness” of online reviews: a cross-cultural study

A critical assessment of consumer reviews: A hybrid NLP-based methodology

“…reviewer belongs to the Top 500 ranks), total reviews posted, reviewer expertise (structured data). These elements of an OCR help to generate many relevant determinants of helpfulness, namely title length [7,53]; title sentiments [7,45,53,69]; review length [7,26,45,46]; review sentiments [26,45,53,69]; review readability [7,22,42,45]; review valence or star-ratings [2,45,46]. Additionally, various attributes of a reviewer play an important role in predicting the helpfulness of online reviews, such as: identity [32], profile [32], actual name/non-anonymity [2,56], credibility [15,56,71] and reputation [15,47,66].…”

“…We accomplished the web-scraping task with the help of an R program and retrieved a total of 30,423 reviews published up to December 2020. Several studies have successfully used Amazon data to examine the helpfulness of online reviews [1,7,14,46,66]. However, the categories have highly different product attributes in terms of product quality evaluability (i.e., search vs experience) and product tangibility (i.e., tangible vs intangible), which significantly affect review helpfulness votes [3,14,21,38,46,64].…”

“…(i) Shannon's Information Theory using entropy 5 package in R; (ii) correlated topic modelling using stm 6 package in R; (iii) text analysis using LIWC [49]; (iv) semantic analysis using lsa 7 package in R. Next, we computed the readability of each review title as: 𝑅𝑒𝑎𝑑𝑎𝑏𝑖𝑙𝑖𝑡𝑦 = 1/ 𝐺𝑢𝑛𝑛𝑖𝑛𝑔𝐹𝑜𝑔𝐼𝑛𝑑𝑒𝑥 where 𝐺𝑢𝑛𝑛𝑖𝑛𝑔𝐹𝑜𝑔𝐼𝑛𝑑𝑒𝑥 = 0.4 * (𝐴𝑆𝐿 + 𝑃𝐶𝑊) ; 𝐴𝑆𝐿 = 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑠𝑒𝑛𝑡𝑒𝑛𝑐𝑒 𝑙𝑒𝑛𝑔𝑡ℎ; and 𝑃𝐻𝑊 = 𝑝𝑒𝑟𝑐𝑒𝑛𝑡𝑎𝑔𝑒 𝑜𝑓 𝑐𝑜𝑚𝑝𝑜𝑢𝑛𝑑 𝑤𝑜𝑟𝑑𝑠 [5]. To measure the superlatives, we computed the number of adjectives in the review title with LIWC [2;5].…”

“…Similarly, we measured the presence of user-provided images of a product in the review as a dummy variable, with a value of one if images were attached alongside a review and zero without them [21,38]. 5 Entropy Package in R: https://cran.r-project.org/web/packages/entropy/entropy.pdf 6 Structural Topic Models Package in R: https://cran.r-project.org/web/packages/stm/stm.pdf 7 Latent Semantic Analysis Package in R: https://cran.r-project.org/web/packages/lsa/lsa.pdf…”

“…Extant literature has reported that review-based attributes [7,22,45,46] and reviewerbased attributes [32,56,66] are suitable for examining the helpfulness of online reviews. However, many challenges remain with these traditional predictors.…”