Persistence of Gender Asymmetries in Book Reviews Within and Across Genres

Abstract

While literary judgments are considered highly subjective or noisy, gender asymmetries are shown in readers' and reviewers' assessments of literature. This study adds genre categories to the examination of literary reviews. By considering both the gender of authors and reviewers, the media type of reviews (newspapers, online blogs), as well as genres, this analysis provides a detailed overview of the Danish review scene and sheds light on structural biases. Analyzing how genres are reviewed in newspapers and blogs, we identify systematic trends that may be attributed to gender biases in reviewer judgments across, as well as within, di昀erent genres and media types. Our results show that the book reviews in Danish media are polarized between the reviewer genders and the two considered media types, upholding additional gender asymmetries in rating.

Keywords

• eol>literary genres
• gender bias
• popularity
• prestige
• literary quality

1. Introduction

While the phenomenon of gender bias in readers’ and reviewers’ perception of the literary ‘quality’ has been observed in multiple contexts14[ , 25 ], the consideration of literary genres provides a more nuanced picture of di昀erential literary judgments across author gend12e,r2[3]. Moreover, since men and women read, review, and publish di昀erent types of literatur14e,[ 23 ], an analysis that does not consider genres will risk a circularity, inferring gender preferences from genre preferences. Therefore, this study considers the additional layer of genres when exploring the potential impact of gender biases on book reviews while examining variations across media types. By examining how genres are reviewed across two di昀erent media types, newspapers and blogs, we seek to assess possible gender biases in reviewers’ judgments across and within genres and media types.

1.1. Gender, prestige, and genre

Various studies have shown that literary reviews are in昀氀uenced by the gender of both reviewers and authors [ 14, 12, 13 ]. Interestingly, these gender biases vary depending on the domains in which assessments occur. In a cross-national analysis of gender di昀erences across the popularhighbrow culture divide, Verboor2d7][ showed that women authors are increasingly present in bestseller lists, while the proportion of women authors remains small among literary award winners.1 Similarly, Lassen, Bizzoni, Peura, Thomsen, and Nielb1o4[] showed how reviews in newspapers display a stronger preference for men writers compared to reviews published in the blogosphere, with a signi昀椀cant preference for women authors. Interpreting this distinction between the institutionally embedded critics in newspapers versus the blog reviewers as ‘grassroots’ critics2[8] as a di昀erence in prestige, allows for an analysis of how women and men authors are perceived di昀erently in the review scene.

In Lassen, Bizzoni, Peura, Thomsen, and Nielbo1[ 4 ], the 昀椀ndings furthermore revealed that preferences for a particular author-gender were linked to the gender majority within each media type. Speci昀椀cally, men dominate as reviewers and authors in newspapers, while women dominate in the same roles in blogs. This intricate interplay between the historical contexts of newspapers and the in昀氀uence of majority preferences highlights the complex nature of social biases within the cultural sphere of literary reviews. As we shall see in the following, genres add further complexities to the study literary judgements.

1.1.1. Genre and implicit quality judgment

Literary genres are hard to de昀椀ne as genre labels are understood as neither fully connected to textual characteristics nor completely extraneous to the6m]. [Genres encompass thematic and formal assessments while also re昀氀ecting implicit quality judgments made by the readers, editors, and critics who assign genres to literary works. Thereby, genres o昀er another layer to the analysis of literary prestige.

In his work, Bourdieu5[ , 4 ] sketched out the literary 昀椀eld in France at the end of the nineteenth century, seeking to assess genres in terms of their placement along two axes: consecration and pro昀椀t [ 7]. Bourdieu distinguishes between types of audiences and considers “consecration by artists, by institutions of the dominant classes, and by popular success” as axes that are more or less mutually exclusi2veH.is mapping of genres o昀ers a practical visualization with a distinction between top-down and bottom-up literary judgments, represented on one side by intellectual and bourgeois audiences, and on the other, by amateur and mass audiences (Fig. 1). Several quantitative studies have built upon Bourdieu’s work, placing literary works along the axes of popularity and prestige2,[ 20 ], and show di昀erences between genres.

Combining aspects of prestige with analysis of gender, Groo8s][has shown that women authors tend to be more o昀琀en labeled or re-labeled into less prestigious genres, such as ‘women’s novel’ over ‘literary novel’. This phenomenon, termed “false categorizing” by R2u1s]s, [leads literary cultures to grant women authors less prestige than men counterparts. One example of 1We use the phrase “women and men authors” instead of the more commonly used “female and male authors” to distinguish the cultural gender (which is examined in this paper) from the biological sex. 2Bourdieu writes: “there are few 昀椀elds [beyond the literary] in which the antagonism between the occupants of the polar positions is more total4”,[p. 46] this is the genre of the ‘literary novel’– a “less popular but more prestigious men-dominated genre” [ 23 ]. Tuchman and Fortin [ 26 ] describe how women writers were ‘edged out’ as the ‘novel’ gained prestige as a literary form – having been shaped into being by women novelists [ 3, 22 ].

Today, the ‘literary novel’ is o昀琀en juxtaposed to ‘genre 昀椀ction’ [ 9 ], and large gender differences persist both in the reader and author distribution across genre divisions, also within ‘genre 昀椀ction.’ For example, Thelwall2[3] shows that there is a substantial gender di昀erence in author gender within most genres: women mainly write – or at least publish – romantic 椀昀ction, children’s literature, and young adult 昀椀ction, whereas men dominate genres such as crime, thriller, and adventure. Evewnithin genres, women and men authors seem to be judged di昀erently [ 24 ]. For example, Higonnet 1[0] reveals gender-based di昀erences in war novels, and research on Bildungsroman1s1[ , 15 ], crime novels 1[], and science 昀椀ction [ 16, 18 ] shows similar patterns, emphasizing the importance of genre in analysis.

This paper aims to examine the e昀ect of gender and genre on reviews published in two di昀erent media types. By mapping out how genres are distributed across media types and the gender of reviewers and authors, our work approaches the following research questions: • RQ1: Which genres are being rated the highest and lowest in each of the two media types (newspapers and blogs)? • RQ2: How does the number of reviews relate to unique book titles within genres? Are there any di昀erences between men and women authors? • RQ3: Within speci昀椀c genres, how is the gender combination between the reviewer and author distributed? Are there rating di昀erences across these combinations? Data set overview No. of reviews Men reviewers Women reviewers Unknown

2.2. Genre retrieval

2. Methods

2.1. Data The dataset used in this work consists of book reviews published in Danish media from 2010 to 2021. The data is retrieved frombog.nu, an online platform collecting book reviews from a variety of sources. For a shared rating scale and gender retrieval, we followed the methods in Lassen, Bizzoni, Peura, Thomsen, and Nielbo1[ 4 ]. In the present work, regional newspapers are aggregated with national newspapers, resulting in 25 newspapers in total written by 620 unique reviewers. Furthermore, blog-like websites are aggregated with blogs, resulting in 203 book blogs dedicated to book reviews. As the bog.nu dataset does not contain IDs for blog reviewers (presumably due to pseudonyms and shared blogs), we do not have the exact number of unique blog reviewers. Similarly, the gender of some blog reviewers is unknown. For an overview of the review, please refer to tab1le. The data retrieved frombog.nu lacks meaningful genre categories, why we collected genre categories elsewhere. We collected two special categories to indicate the canonicity of titles (national or international): for the category, ‘Danish classics’, we relied on the reference work Danish poets in the 20th century [ 17 ] and the Danish Cultural Canon1[ 9 ], retrieving a list of 160 unique authors, of whom 113 were extant in the bog.nu dataset. For ‘International classics,’ we made use of the Open Syllabusby selecting the 1,000 most assigned authors in ‘English Literature’ syllabi resulting in a list of 502 unique authors, of whom 77 were extant in the bog.nu dataset. For the remaining genres, we used the ISBN system, scraping the DK-5 or ‘theme codes’ connected to each ISBN on the large online bookshoSpaxo, which resulted in 31 unique genre categories. A昀琀er manual inspection of the genre codes, we merged similar genres, preferring more overarching categories in the DK-5 syste3mW. e thus ended up with 18 unique genres. See Fig.2 for a visual overview of the genre retrieval pipeline. 3See all categories ath:ttps://ns.editeur.org 12441 5968 1953 3649 977 1997 2055 1498 1197 1240 1235 1010 941 904 394 394 349 173 6625 4.28 4.36 2.03 4.31 1.85 3.86 4.03 3.52 3.02 3.90 4.06 3.88 4.07 4.66 2.54 3.49 3.92 2.98 2.36 2.3. Model In order to estimate the relative e昀ect of genre on the reviewer-assigned rating (six-point scale), we 昀椀t the following linear model: rating = 0 + genre genre + with the null model tha tgenre = 0. Where is the rating of review , is the predictor value (genre) of review , represents unknown parameters, and is the error terms. A linear model is 昀椀tted for both blogs and newspapers, respectively.

To estimate the relative e昀ect of author and reviewer gender on reviewer-assigned ratings, we have analysed each genre separately4.To do so, we have 昀椀tted the following linear model: rating = 0 + author gender author gender+ reviewer gender reviewer gender + with the null hypothesis that author gender= reviewer gender = 0. A linear model is 昀椀tted for both blogs and newspapers, respectively. For models 昀椀tted on data for newspapers, a random e昀ect for each reviewer is added. (1) (2)

3. Results

3.1. Distributions

3.1.1. Genre distribution

Looking at the total number of reviews (See Tabl2efor the number of reviews, and Fig3. for percentage distribution), we 昀椀nd that beyond ‘Literary 昀椀ction’ and books without a known category, the genres ‘Crimes & Mysteries’, ‘Young Adult’ (YA), and ‘Fiction for Kids’ are the most reviewed genres. Looking at reviews in newspapers speci昀椀cally, the most reviewed genres are ‘Crime & Mystery’, ‘Danish classics’, and ‘Poetry’. In blog reviews, the most reviewed genres are ‘Crime & Mystery’, ‘YA’, and ‘Romantic Fiction’. Looking at the review ratio of each genre (See Table2), we see that it is highest for the genres ‘Biographical 昀椀ction’, ‘Crime & mystery’, ‘YA’ and smallest for ‘Books for small children’, ‘Fiction for children’, and books without a known category.

3.1.2. Gender distribution

The distribution of the total number of titles by women and men authors within each genre in the bog.nu dataset show patterns similar to those in Thelwal2l3][: titles by women authors are predominant in genres like ‘Romantic 昀椀ction’, ‘Erotic 昀椀ction’, and ‘YA’. Whereas men authors are more represented in ‘Graphic novels’, ‘Science 昀椀ction’, and ‘Classics’ (both Danish and International). opt for this approach due to interpretability, as it would otherwise involve dealing∼w8it0hparameters. 4A common modelling approach is to include a genre variable rather than working with data subsets. However, we

The di昀erences in author gender across genre is further explicit when we nuance the data and look at reviews in newspapers and blogs, respectively. In newspapers, reviews of mendominated genres are more present, whereas books of women-dominated genres are reviewed more in blogs.

Considering the di昀erence between the proportion of reviews of titles by women/men authors and the proportion of actual titles by women/men authors in each genre, we 昀椀nd an opposite trend in blogs compared to newspapers: in blogs, titles by women authors tend to be reviewed disproportionatelmyore in relation to actual number of titles by women authors, while they are reviewed disproportionatelleyss in newspapers in relation to the number of titles by women (Fig. 4). In newspapers, the percentage of reviews of women authors is less than or equal to the percentage of books that are written by women authors in almost every genre (i.e., the light-green bars tend to be longer than the dark-green bars in F4i,gr.ight). We 椀昀nd that the proportion between reviews and titles of women authors is most unequal for the genres ‘Horror’, ‘Danish classics’, and ‘Poetry’. Where the former is a small genre (T2a)b,tlhee two others are connected to high prestige2[0](cf. Fig. 7). In blogs, the opposite seems to be the case, where men authors consistently reviewed disproportionatelelsys (i.e., the red bar in line Fig.4 tends to be longer than the pink). In sum, we 昀椀nd opposite tendencies when looking at the relation between reviews and no. of titles in newspapers in contrast to blogs. Note that this bears no relation tohow women/men authors are reviewed. 3.2. Ratio As earlier studies have shown that men and women reviewers mostly review authors of their own gender [ 14 ], we have here examined the gender ratio of the reviews of each genre. Similarly, for media type, we have calculated a media ratio between blogs and newspapers in order to see where each genre is mostly reviewed.

Plotting the media ratio on th e -axis and the gender ratio on the -axis, we obtain an overview of where each genre is reviewed. We observe that all 18 genres show a high density of reviews in the corners, representing men reviewers in newspapers and women reviewers in blogs. Distinguishing between books by men and women authors, we furthermore see that reviewers of both genders mostly review books written by their own gender, aligning with the results in Lassen, Bizzoni, Peura, Thomsen, and Nielbo14[]. In the following, we focus on two categories, ‘Crime & Mystery’ and ‘International classics’, presenting their full results. Beyond the generic categories (‘Literary 昀椀ction’ and ‘Unknown’), ‘Crimes & Mysteries’ is the most reviewed genre in both blogs and newspapers, while ‘International classics’ holds high prestige (see Fig. 7).

Fig. 5 shows point plots of where the two genres are reviewed. We see that newspaper reviews of ‘Crime & Mystery’ are mostly of men authors with a division on reviewer gender. In other words, reviewers are predominantly reviewing crime novels written by an author of their own gender. In blogs, we see that most reviewers are women, con昀椀rming the 昀椀ndings in Lassen, Bizzoni, Peura, Thomsen, and Nielbo1[ 4 ] and that these women reviewers show larger diversity in author gender preference compared to women reviewers in newspapers. In other words, women reviewers in blogs review crime novels by both genders.

For the genre ‘International classics’, we observe a split on the diagonal, indicating that men review in newspapers and women review in blogs. The newspaper reviews are mostly of men authors, and the few reviews of women authors are mostly reviewed by women reviewers. On blogs, we see a similar preference for same-gender authors.

For newspaper reviews, 昀椀tting (grade of review) in the model (1) above with a mixed-e昀ect model with reviewer ID as a random e昀ect. This allows us to see the e昀ect of genre and but also the random variability introduced by di昀erent reviewers, which is shown to be 0.089 with SD = 0.012, which we interpret as relatively small on point-six scale.

3.3. Rating

3.3.1. Newspapers

coef 3.3.2. Blogs As we do not have reviewer ID for blog reviews, we 昀椀 t (grade of review) in the model (1) above with a 昀椀xed-e昀ect model (ordinary least squares), resulting in statistically signi昀椀cant rating di昀erences in 9 of the 18 di昀erent genre categories.

coef

3.3.3. E昀ect of gender within genre

We examine the e昀ect of reviewer-author gender combinations for reviews of each genre in newspapers and blogs, respectively. We limit this section to include the results on the genres ‘Crime & mystery’ and ‘International classics’.

For reviews of crime novels (see le昀琀-hand side of Fig.6), we 昀椀t (grade of review) in the model (2) above with ordinary least squares (OLS) for blog reviews and a mixed-e昀ect model for newspaper reviews. For newspaper reviews we observe a larger span in the ratings with statistically signi昀椀cant rating di昀erences for the gender combinations: woman-women and woman-man, with < 0.05 . (Group Var = 0.101, Std. error = 0.036 for random e昀ect for reviewers). For reviews on blogs, on the other hand, reviews of crime novels show no statistically signi昀椀cant di昀erences between the gender combinations of reviewer and author.

We performed the same analysis for reviews of ‘International classics’ in newspapers and blogs, respectively, by 昀椀tting (grade of review) in the model (2) above with ordinary least squares (OLS) for blog reviews and a mixed-e昀ect model for newspaper reviews (see righthand side of Fig.6). For blog reviews, there are statistically signi昀椀cant di昀erences between the reviewer-author gender combinations: woman-woman and woman-man. Newspapers reviews show statistically signi昀椀cant di昀erences between the gender combinations woman-woman and man-man (Group Var = 0.135, Std. error = 0.051 for random e昀ect for reviewers). Hence, 昀椀nd that newspaper reviews mirror the results found on all reviews in newspap1e4r]s, w[here and women reviewers rate women authors the lowest and men reviewers the highest - and the opposite for blog reviews. We furthermore observe that the genre ‘International classics’ are rated similarly in the two media types.

This indicates that not only does the e昀ect of gender di昀er between the media types, but also across di昀erent genre categories.

4. Discussion

The work presented in this paper shows that the cultural sphere of book reviews constitutes a complex landscape characterized by intricate interactions between gender dynamics and genre preferences. Our analysis shows that the provided genre categories do not fully elucidate the found gender asymmetries in book reviews. Instead, they add another crucial layer to a nuanced understanding of structural biases in the Danish review scene, split between two genders and the two considered media types. Moreover, sectio3n.2 shows that both genders predominantly review authors of their own gender – men reviewing in newspapers and women in blogs. This trend is partially explained by a media-type breakdown of reviews in S3e.1c.2, showing that genres dominated by women authors receive more attention in blogs, while genres dominated by men authors receive more coverage in newspapers.

4.1. Prestige and popularity - mapping the Danish review scene

Following the heuristic dichotomy of popularity and prestige as in Bourd5i]e,uw[e provide a similar map of the cultural 昀椀eld of book reviews. Popularity, represented on t-haexis, is derived from average ratings, while prestige, shown on th-eaxis, is calculated by the media ratio of each genre’s review5s.

In Fig. 7, we see that ‘Danish classics’, ‘Poetry’, and ‘Graphic novels’ are of high prestige, whereas the genres ‘YA’, ‘Fantasy’, and ‘Horror’ are of low prestige, aligning with 昀椀ndings in Porter [ 20 ]. The colors of the circles represent the author distribution, with dark colors corresponding to mostly women authors, measured on the number of published reviews. Hence, women dominate in the genres, which are low-prestige genres (mostly reviewed in blogs). Men, on the other hand, dominate in genre located at the high-prestige end of the spectrum (mostly reviewed in newspapers). This corresponds to the popular-highbrow culture divide examined 5As reviews in blogs give, on average, higher ratings than reviews published in newspapers, there is a skewness towards high popularity in low-prestige genres. by Verboord 2[7].

Rating of speci昀椀c genres di昀ers across the two media types. The poetry genre, which is rated the lowest in blogs∼( 0.5 points lower than intercept), is mainly reviewed in newspapers resulting in Prestige > 0.8 in Fig7. . Similarly, genres that are rated low in newspapers, such as ‘Thriller & suspense’ and ‘Romantic 昀椀ction’, are reviewed more in blogs – however, they are still reviewed in newspapers. Blog reviewers tend to avoid genres they dislike, while newspaper reviewers continue to review low-rated genres, possibly in昀氀uenced by editorial choices in newspapers. This disparity in grading behaviours between the two media types, as noted in Lassen, Bizzoni, Peura, Thomsen, and Nielbo1[ 4 ], may be attributed to these factors.

4.2. Outlook

In conclusion, we see that gender di昀erences vary from genre to genre. Further work is still needed to model the full landscape of the Danish review scene (e.g., a Bayesian model including all parameters), but from the presented work we see that when the two media types align with the rating of a genre, for instance, ‘International classics’ and ’Graphic novels’, there are di昀erent gender disparities at play compared to genres with rating disagreement between the two media types. Analysing genres individually, we see the same pattern of structural biases on a small scale as found on a large scal1e4[]: low ratings of women authors by women reviewers and higher ratings of men authors by men reviewers. This highlights the persistence of gender biases even within smaller subcultures as well as in larger social spheres, emphasising the structural aspect of the found asymmetries. [13] C. W. Koolen. “Reading beyond the female: the relationship between perception of author gender and literary quality”. PhD thesis. Amsterdam: Institute for Logic, Language and Computation, Universiteit van Amsterdam, 2018.

5. Online Resources

References

[1] [2] [3] [7] [8] [10] [11] A. Alacovska . “ The gendering power of genres: How female Scandinavian crime 昀椀ction writers experience professional authorship” . IOn:rganization 24.3 ( 2017 ), pp. 377 - 396 .

doi: 10 .1177/1350508416687766.

Canon/archive: large-scale dynamics in the literary 昀椀eld . Pamphlets of the Stanford Literary Lab 11 . Stanford, CA, USA: Stanford Literary Lab, 2016 . urhlt:tps://litlab.stanfor d. edu/LiteraryLabPamphlet11 .p d.f N. Armstrong . Desire and domestic 昀椀ction: a political history of the novel . New York, NY, USA: Oxford University Press, 1987 .

[4] P. Bourdieu . The 昀椀eld of cultural production: essays on art and literature . Ed. by R. Johnson . New York, NY, USA: Columbia University Press, 1993 .

[5] P. Bourdieu . The rules of art: Genesis and structure of the literary 昀椀eld . Stanford, CA, USA: Stanford University Press, 1996 .

[6] J. Derrida . “ The Law of Genre” . Trans . by A. Ronell . ICnr:itical Inquiry 7.1 ( 1980 ), pp. 55 - 81 .

A. Dunst . “ How we read comics now: Literary studies, computational criticism, and the rise of the graphic novel” . InM:FS Modern Fiction Studies 67.4 ( 2021 ), pp. 758 - 784 . doi: 10 .1353/mfs. 2021 . 0040 .

M. Groos . “Wie schrij昀琀 die blij昀琀? Schrijfsters in de literaire kritiek van nu”. InT:ijdschri昀琀 voor Genderstudies 3.3 ( 2000 ).

[9] S. Gupta . “ On Mapping Genre: Literary Fiction/Genre Fiction and Globalization Processes” . In: Globalizing literary genres: Literature, history, modernity. Ed. by J. Habjan and F. Imlinger . Oxfordshire, England, UK: Routledge, 2015 , pp. 213 - 227 .

M. R. Higonnet . Borderwork: Feminist Engagements with Comparative Literature . Ithaca, NY , USA: Cornell University Press, 2018 . do1i: 0 .1353/book.58027.

M. Hirsch . “ Jane's Family Romances” . In:Borderwork: Feminist Engagements with Comparative Literature . Ed. by M. Higonnet . Ithaca, NY , USA: Cornell University Press, 1994 , pp. 162 - 186 .

[12] C. Koolen , K. van Dalen-Oskam , A. van Cranenburgh , and E. Nagelhout. “ Literary quality in the eye of the Dutch reader: The National Reader Survey” . IPno:etics 79 ( 2020 ). doi: 10 .1016/j.poetic. 2020 . 101439 .

[14] I. M. S. Lassen , Y. Bizzoni , T. Peura , M. R. Thomsen , and K. L. Nielbo . “ Reviewer Preferences and Gender Disparities in Aesthetic Judgments” . InC:EUR Workshop Proceedings 3290 ( 2022 ), pp. 280 - 290 .

[15] C. Lazzaro-Weis . “The Female “Bildungsroman”: Calling It into Question” . NInW:SA Journal 2.1 ( 1990 ), pp. 16 - 34 . doi: 10 .1017/9781316479926.009.

[16] B. Lovell . “ Science 昀椀ction's women problem” . In:Creative Writing Scholars on the Publishing Trade . Ed. by S. Meekings and M. Moore . Oxfordshire, England, UK: Routledge, 2021 , pp. 68 - 83 .

[17] A.-M. Mai . Danske Digtere - i det 20. århundrede. Website . 2000 . url: https://www.dansk edigtere.sdu.dk./

[18] H. Merrick . “ Gender in science 昀椀ction” . In: The Cambridge Companion to Science Fiction. Ed. by E. James and F. Mendlesohn . Cambridge, England, UK: Cambridge University Press, 2003 , pp. 241 - 252 . doi: 10 .1017/ccol0521816262. 019 .

[19] F. H. Mortensen . “Litteratur: Bidrag til” Kulturkanon””.KInu:lturkontakten: Saernummer. Copenhagen, Denmark: Kulturministeriet, 2006 , pp. 40 - 47 .

[20] J. Porter . Popularity/prestige: A new canon . Pamphlets of the Stanford Literary Lab 17 . Stanford, CA, USA: Stanford Literary Lab, 2018 .

[21] J. Russ . How to Suppress Women's Writing . Austin, Texas, USA: University of Texas Press, 1983 .

[22] M. D. Stetz . “ Gender and Genre” . In:NOVEL: A Forum on Fiction 25.1 ( 1991 ). Ed. by A. L. Ardis , pp. 108 - 110 . doi: 10 .2307/1345669.

[23] M. Thelwall . “ Book genre and author gender: Romance>Paranormal-Romance to Autobiography>Memoir” . In:Journal of the Association for Information Science and Technology 68.5 ( 2017 ), pp. 1212 - 1223 . doi: 10 .1002/asi.23768.

[24] M. Thelwall . “ Reader and author gender and genre in Goodreads”J.oInu:rnal of Librarianship and Information Science 51.2 ( 2019 ), pp. 403 - 430 . doi: 10 .1177/0961000617709061.

[25] S. Touileb , L. Øvrelid , and E. Velldal. “ Gender and sentiment, critics and authors: a dataset of Norwegian book reviews” . InP:roceedings of the Second Workshop on Gender Bias in Natural Language Processing . Barcelona, Spain (Online): Association for Computational Linguistics , 2020 , pp. 125 - 138 . url:https://aclanthology.org/ 2020 .gebnlp- 1 . 1. 1

[26] G. Tuchman and N. E. Fortin . Edging women out: Victorian novelists, publishers and social change . Vol. 13 . Oxfordshire , England, UK: Routledge, 2012 .

[27] M. Verboord . “ Female bestsellers: A cross-national study of gender inequality and the popular-highbrow culture divide in 昀椀ction book production , 1960 - 2009 ”. IEnu:ropean Journal of Communication 27.4 ( 2012 ), pp. 395 - 409 . doi: 10 .1177/026732311245943.

[28] M. Verboord . “ The legitimacy of book critics in the age of the Internet and omnivorousness: Expert critics, Internet critics and peer critics in Flanders and the Netherlands” . In: European Sociological Review 26.6 ( 2010 ), pp. 623 - 637 . doi: 10 .1093/esr/jcp039.

See https://zenodo.org/records/1001126f9or code.