Toward Understanding Book Rating Divergence across Platforms and Editions:A Preliminary Case Study

1. Introduction

Online book ratings on social reviewing platforms, such as Goodreads and Amazon, have been investigated across disciplines for illuminating reader preferences, social impacts of books, and book sales, etc. (Dimitrov et al., 2015; Maity et al., 2019; Murray, 2021; Wang et al., 2018; Zuccala et al., 2015). However, despite the availability of distinct web pages for rating different editions of the same works on social reviewing platforms, prior DH studies used either the rating of the most rated single edition (Hu, LeBlanc, et al., 2023) or the aggregated rating across editions to represent the works’ reputation online (Walsh & Antoniak, 2021). Such single-faceted conventional practice, while being effective for large-scale data analyses, unavoidably simplifies the divergent ratings towards different editions of works to a single rating. This might downplay the significance of book editions in reception (Garritzen, 2018; Genette, 1997) and may cause analytical oversights. For instance, a book’s original publication and translated versions may receive vastly different ratings across readerships and platforms. These differences cannot be reflected by singular ratings and have not been adequately investigated in existing DH research on online book reviews.   

2. Datasets

Reflecting on this limitation, we have undertaken an exploratory study that posits book ratings as multi-faceted and dynamic indicators of books’ online reputation, and advocates for investigating the complexities and nuances of book ratings across platforms and editions. Instead of collecting ratings of massive collections of books, we have opted to explore a small set of “Classics” (Hu, Underwood, et al., 2023) that have a wider reception history across cultures and have been rated across platforms and book editions respectively. These 23 books, listed in Table 1, were identified as “User-Elected Classics” across Goodreads and Douban (the two of the largest book reviewing platforms based in the U.S. and China) by our prior research (Hu, Underwood, et al., 2023). Data from these two platforms not only provides cross-cultural perspectives (Hu, LeBlanc, et al., 2023) but also offers comparisons of rating systems: Goodreads aggregates ratings for the same work across editions ¹ and presents the same aggregated rating on the linked book webpages, while Douban maintains ratings given to different editions of the same work separately. 

For our case study, we took One Hundred Years of Solitude (OHYOS), one of the 23 “Classics.” A search for its book titles in English, Spanish, and Chinese (two most popular translations) across Douban and Goodreads uncovered 39 distinct ratings items. On Goodreads, as shown in Figure 1, webpages for three different OHYOS editions in Spanish, English, and Chinese showed different metadata, but the same aggregated rating (4.11 out of 5, rated by 944,741 users). On Douban, however, as Figure 2 presents, each OHYOS edition was rated individually, presenting different ratings.

Table 1

The 23 Classics across Goodreads and Douban

Figure 1 

Webpages for Three Editions of One Hundred Years of Solitude

Note. The three editions are in English, Spanish, and Chinese, showing the same aggregate rating on Goodreads ( One Hundred Years of Solitude, 2023; Cien Años de Soledad, 2023; 百年孤独, 2023)

Figure 2 

Different Editions of One Hundred Years of Solitude 

Note. The editions are searched in English, Spanish, and Chinese, and have varied ratings on Douban ( Douban Search “Cien Años de Soledad,” 2023; Douban Search “One Hundred Years of Solitude,” 2023; Douban Search “ 百年孤独 ,” 2023)

3. Results

Figure 3 visualizes the ratings (x-axis) and numbers of ratings received (y-axis) that showed up on different editions’ webpages across Goodreads (7 webpages) and Douabn (32 webpages), retrieved by OHYOS titles in three languages (see Appendix Table 1 for more details). This figure represents the 39 vastly divergent ratings of OHYOS across platforms and editions that are customarily aggregated into a single rating in most book rating studies. Note the strong divergence across platforms (blue v. orange) and amongst the many different editions reviewed on Douban.

Figure 3

Distribution of Book Ratings and Number of Ratings Received of 39 Editions of OHYOS across Douban and Goodreads

Note. The results are limited to true positive matches returned by each query on the first page.

4. Ongoing and Future Work

The strong signal of ratings divergence apparent in the OHYOS case study has led us to expand our analysis across the other 22 titles in our “Classics List.” To explore this phenomenon, for each work in the “Classics list” above, we are collecting ratings for all their various editions across the two platforms in order to understand:

1) How online book rating might diverge across platforms and editions; 

2) What are the nuances in ratings that might have been removed by the “one book one rating” practice in DH research; 

3) What are the potential edges or pitfalls of using one aggregated rating vs. multiple varied book ratings in DH research? 

Through our attempt to answer these questions, we aim to reimagine and reinvent current DH investigations toward a more critical and comprehensive understanding of book ratings online.   

5. See Also

39 Different Editions of One Hundred Years of Solitude Retrieved across Goodreads and Douban

Note. For more information regarding how we aggregated the 2-10 scale in Douban and 1-5 scale in Goodreads, see Hu, LeBlanc, et al., 2023.