Gendered Hate Speech in Swiss WhatsApp Messages

This paper provides a first manual analysis of the use of 14 gender-specific swear words in Swiss WhatsApp messages in (Swiss) German, French and Italian. We will show that gender-specific swear words only concern non-participants of the discourse when used offensively, in our data. When such items are directly addressed to the interlocutor, they are rather used ironically or even hypocoristically. Furthermore, we will show that gendered hate speech in this context is not a matter of mutual discrimination by the two sexes, but that Swiss WhatsApp users also refer to people of their own sex using gender-specific swear words.


Introduction
The interest in gender equity and the prosecution of verbal, physical or mental harassment related to gender and/or sexual orientation has been increasing over the past several years. Especially the verbal aspect of gender-based discrimination, namely (gendered) hate speech, is regarded as particularly challenging in terms of detection, reporting and prosecution, since it lacks an overall accepted definition (cf. Alkiviadou 2018). For example, the Council of Europe's Committee of Ministers does not mention hatred based on gender differences or sexual orientation explicitly as instances of hate speech (cf. Alkiviadou 2018, 23): [Hate speech is] understood as covering all forms of expression which spread, incite, promote or justify racial hatred, xenophobia, anti-Semitism or other forms of hatred based on intolerance, including: intolerance expressed by aggressive nationalism and ethnocentrism, discrimination and hostility against minorities, migrants and people of immigrant origin. (Recommendation (97)20, Council of Europe, Committee of Ministers 1 ) One of the main reasons why gendered hate speech seems to constitute a complex and to date underresearched category of verbal harassment is that it is difficult to discern whether hateful expressions are really used to harm somebody, or if, instead, they are intended to be understood ironically or sarcastically. This aspect is especially demanding for the interpretation of written data where we lack prosodic information and facial expressions, as will be shown below. This is by no means exclusively a potential impediment for the researcher, but it can also have a confusing impact on the actual receiver of the message. We seek nevertheless to present in this chapter a first manual quantitative, cross-linguistic analysis of gendered hate speech in the multilingual Swiss WhatsApp corpus (Stark, Ueberwasser, Göhring 2014-20), providing data from a nowadays ubiquitous form of computer-mediated communication, WhatsApp (Ueberwasser, Stark 2017). To do so, we will analyse a determined set of lexical items such as 'bitch' or 'wanker', which can potentially be classified as instances of gendered hate speech, in their pragmatic use in the Swiss What-sApp corpus. We will also have a look at the gender of the message's author. Thus, in our pilot study 2 we are addressing the following research questions: • Which lexical items characterize gendered hate speech in Swiss WhatsApp messages? • What are the functions of these lexical items? • Are there any gender-specific differences? 1 https://search.coe.int/cm/Pages/result_details.aspx?ObjectID =0900001680505d5b.
2 In this chapter we present the results of a pilot study carried out in Summer 2018 as a preparation for a large citizen science project started in spring 2019 in collaboration with the Citizen Science Centre of the University of Zurich and ETH Zurich.
In order to answer these questions, we will show an analysis of data from three languages ([Swiss] German, French and Italian).
The chapter will be structured as follows: first, we will briefly present the state of the art with regard to (gendered) hate speech. The methodology of our pilot study and how we tried to diminish the challenges of dealing with written data (as mentioned above) will be explained in section 3, followed by the main results (section 4), which will be discussed in section 5. In section 6, we will present our conclusions.

2
Gender and Hate Speech. State of the Art The relationship between gender and language has always been considered complex, let alone the relationship between gender and a very specific form of language use, namely hate speech (cf. Coates 1992;Crawford 1995;Stapleton 2003). In our study, we address the factor of gender in a twofold way: we focus on the relationship between the sex of a speaker/writer and what is known as gendered hate speech, namely the frequency in use of gender-specific swear words in Swiss WhatsApp messages and their functions. To do so, it is essential to first define what can be understood as gendered hate speech, given the fact that the aspect of gender (as a reason for being attacked) is missing in many official international documents that deal with discrimination and the spread of hatred (cf. Alkiviadou 2018). In our study, we define gendered hate speech as follows: Gendered hate speech encompasses expressions which are used to harm, insult, disrespect, or discriminate an individual or a certain group of people on the basis of gender or sexual orientation. This includes derogatory expressions referring to the sexual anatomy, gender-specific swear words, and expressions to designate members of the LGBTQ-community.
Numerous studies have been carried out on a more general level, comparing the use of swear words for both male and female speakers (Risch 1987;de Klerk 1988de Klerk , 1991de Klerk , 1992Stapleton 2003, among others) and some of those studies also addressed the use of genderspecific swear words and their evaluation by men and women (cf. de Klerk 1991Klerk , 1992Klerk , 1997. Most of them started from a very similar hypothesis, which is primarily based on the general belief that women tend to use less 'bad language' than men. Secondly, some of the first studies carried out in this area of interest (cf. Jespersen 1922;Trudgill 1974;Lakoff 1975) reinforced this stereotype by qualifying women as "avoiders" of swear words because they pay more attention to linguistic norms and values such as politeness and status (Stapleton 2003;de Klerk 1992). Also, breaking so-called "linguistic taboos" (de Klerk 1991) is often associated with "masculinity and toughness" (Hughes 1992, 291). Chetty and Alathur, introduced the term of "sexist hate speech" (2018, to refer to the concept of gendered hate speech, which, according to them, is generally targeted towards women or girls. Regarding the gender-related "imbalance" in the frequency in use of swear words, Stapleton (2003) also argued that men use terms referring to the female sex organs on a regular basis, whereas women tend to avoid the same terms (see also Fine, Johnson 1984). However, the empirical data analysed in the above-mentioned studies provided -at least partially -counterevidence to the widespread stereotypes, which were shown to be basically overgeneralizations (Hughes 1992). Many researchers found that the frequency in use of swear words by women is not significantly lower than the frequency with which men use swear words. Indeed, the use of so-called 'bad language' seems to be regularly used by both sexes (Stapleton 2003;de Klerk 1988de Klerk , 1991de Klerk , 1992. Referring concretely to gendered hate speech, Risch (1987) showed that women do have and use a vast range of derogatory expressions to refer to men. Also, de Klerk (1992) was able to negate the hypothesis of "lexical bias" in a study on gendered slang, which claims that the idea that men tend to use more "bad words" referring to women is based on the fact that there are more gender-specific swear words or "slang words" referring to women in general. De Klerk (1992) showed, for example, that women never were short of "slang words" to refer to males. Moreover, de Klerk also claimed that, in connection with swearing and the use of slang, the role of solidarity seems to be essential for women: Females, it would appear, are not striving for standard prestigious speech (Trudgill 1972) but are striving to use what their peers are using. It would seem that males and females alike gain solidarity from using nonstandard words and that females (certainly young ones) are not necessarily as linguistically conservative as current literature would have us believe. (1992, 286) Sutton's (1995) study on slang words for women reported another interesting conclusion: several (female) informants revealed that the interpretations of the expressions 'ho' and 'bitch' were context-bound. That means that when women refer to each other as 'ho' or 'bitch', it can even be an expression of endearment, whereas when men use these terms, it is always considered offensive by women. Sutton also claimed that "slang used by young women to address one another is part of their search for identity as individuals and as a group (women) in a male-dominated world" (1995,289) and that women who use slang words do not actually talk "like men", but they tend to imitate other women who stand out for not fulfilling the typical "female stereotypes". Thus, women's swearing seems to be restricted to a limited number of contexts. Stapleton (2003, 23) argued, based on Risch's (1987) and Hughes' (1992) statements, that women's swearing presupposes an ambience of intimacy or trust more than for men. This aspect is particularly interesting if we take into account data from the new media such as Facebook or text messages, as hate speech seems to become more and more prevalent in the internet, more specifically in social media (Guiora, Park 2017), which means that hate speech seems to shift from the more private realm towards a public space.
However, the data in which we are interested in our study belong to the context of private conversation. WhatsApp communication itself, as a relatively new form of computer-mediated communication, has not received a lot of attention in linguistic research so far (Ueberwasser, Stark 2017), never mind regarding a topic such as hate speech. WhatsApp is a smartphone application for written communication and has become the widely preferred successor of 'normal' text messages as it provides a platform for the free exchange of written messages, audio or visual material for which only an active connection to internet is needed (cf. Dürscheid, Frick 2014). If both interlocutors are online, it is even possible to have some sort of "real-time communication" (Dürscheid, Frick 2014, 167). Another feature which is frequently used in WhatsApp communication is the rich catalogue of emoticons and emojis, which often assume an expressive or emotive function, but they are also used on a propositional level as a substitute for a written word (Dürscheid, Frick 2014, 173-4).

Methodology
Our study is based on a corpus resulting from a collection of approximately 600 authentic WhatsApp chats which was carried out in summer 2014 in Switzerland. The chats comprise circa 750,000 anonymized messages in (Swiss) German, French, Italian and Rhaeto-Romance, totalling over 5 million tokens. There is a considerable difference in the amount of data in the different sub-corpora: there are over 3 million tokens in the German and Swiss German sub-corpora, whereas in the French and Italian sub-corpora there are only 1.3 million and 340,019 tokens respectively. The chats were browsed in AN-NIS, a search and visualization architecture which allows us to create and use different kinds of search options (Ueberwasser, Stark 2017).
For the pilot study, the following 14 gender-specific lexical items, which all denote sexual anatomy, sexual orientation or professions related to the world of sex-workers and do barely exhibit graphic variants, were analysed using the ANNIS query tool: Table 1 Lexical items taken into account in the pilot study

Language
Item Translation 'son of a bitch' * This item is written Wichser in Standard German, but we only find this variant in our data.
For the analysis, the following parameters were taken into account: a. the language of the message ((Swiss) German, French, Italian), b. the gender of the author of the message and c. the intended meaning of the message split up into three modalities of context: clearly disrespectful (offense), clearly not disrespectful (endearment/irony), and unclear. The last of these three categories is, of course, problematic, as a lot of messages containing a gender-specific swear word were attributed to this category. This is due to the fact mentioned above that the detection of hate speech is a tricky task because we lack important para-and non-verbal information such as prosody or facial expressions, which could give us a hint as to whether an expression is used ironically or seriously.
Despite this shortcoming, informal written data, and WhatsApp messages in particular, provide a series of important contextual elements to discriminate a clearly disrespectful from a clearly not disrespectful (ironic or even hypocoristic or endearing) meaning. To discern a clearly disrespectful meaning, for instance, elements such as modifiers like pejorative adjectives ((1) and (2)), demonstrative determiners ((3) and (4)) and intensifiers ( (5) and (6)) can be helpful: (1) ger. Die isch en hässlichi Nutte 'She is an ugly slut.' fr. C'est la pire pute 'She is the worst slut.' (3) ger. ih hasse so lehrer ih ha gaar nüt gseit u die bitch faat afa liire.. 'I hate those teachers, I said nothing, and that bitch begins to drone…' (4) fr. Il y a de ces poules dans la classe Passerelle 'There are those sluts in the 'Passerelle'-class.' On the other hand, the use of emojis and emoticons can provide a clue to intended irony and sarcasm, as we can see in the example in Figure 1, where the laughing emoji relativizes the swearword fotze (cunt): 3

Karma Is a Bitch
3 Translation: 'You fucking cunt, take me along!'.
4 It is certainly true that also interjections, discourse markers and quotations containing a swear word can have a distressing impact on the reader, but for this pilot study we only considered swear words referring to human referents.
Altogether, we found only 2,315 occurrences of the 14 lexical items in our relatively large corpus of ~ 5.5 million tokens. Among those occurrences, almost 80% were excluded because they were false positives (in the sense described above). After having discarded false positives, we worked on 502 examples in three languages. As a second step, we analysed the data according to the three above-mentioned modalities of context by taking into account the previous and the following message and elements such as those described above to clearly distinguish disrespectful meanings from ironic or hypocoristic meanings.

Results
Of the three categories of functions we determined for our analysis, we will focus on the clearly disrespectful use of gender-specific swear words in what follows. This overview shows that, in general, words referring to the concept of 'prostitute' ((swiss) ger. Bitch, Nutte; fr. salope, pute) (61-86%) and the German word for 'wanker' (73%) were most often used in an offensive and clearly disrespectful context. However, the French polysemous word poule, which refers to the concept of 'prostitute' or 'mistress', but also to the concept of '(girl)friend' (especially when it is accompanied by a possessive as in fr. ma poule), 5 was only used offensively in 5% of cases. The pragmatic function of this term was unclear in 69% of the occurrences. Surprisingly, only 56% of the occurrences of the word denoting the female sexual anatomy ((swiss-) ger. Futz/Fotze) appeared in a derogatory context. Figure 2 shows the frequency of the terms denoting women in an offensive context. Regarding the swear words for men [ Fig. 3], the terms denoting a male homosexual seem to be very common in everyday conversation and are not used as instances of hate speech (anymore) as they appeared in an offensive context in only 8-27% of the occurrences. On the contrary, we find many contexts where those terms were used with an ironic or even hypocoristic meaning. The term Homo (homosexual) always refers to the (male) interlocutor when it is used in an ironic context and to a third (male) person when used as hate speech. The term gay appears in the German and the French subcorpora. In each of them, there is only one occurrence of the term in an offensive context (1 out of 5 in the German corpus; 1 out of 19 in the French corpus). The French term referring to a female homosexual (gouine), in contrast, was not included in the discussion because it only appeared once in the corpus and its function was not clearly classifiable.
The German term for 'wanker' (Wichser) is used with a negative, pejorative meaning in 73% of the occurrences. The Italian word figlio di puttana (son of a bitch) was not taken into account in the presentation of the results because it only occurred twice in our corpus (it was used in a derogatory context in both cases and referred in both cases to a male third person who was not part of the conversation). The differences between men and women in their active use of swear words in a disrespectful manner, described as highly significant in the previous literature, could not be confirmed by our data. The following two figures give the distribution of males and females who used the respective swear words denoting women [ Fig. 4] and men [Fig. 5] in an offensive context. The first comment that can be made is that the frequency with which women use those words offensively is considerably higher than would be expected given the claims in previous studies. For instance, the use of the English word Bitch (bitch) in the German corpus constitutes the big exception in our data: it is the only term which is more often used by female writers (83%) than by male writers (17%). Another interesting and rather surprising finding is that ger. Futz/ Fotze (cunt) is almost equally often used by men (53%) and women (47%), although it has been claimed that most women tend to avoid terms referring to the female sexual anatomy (cf. Stapleton 2003;Fine, Johnson 1984).
If we compare the difference in use between the two genders for the terms referring to men, we can see that, interestingly, the major-

Wichser Schwuchtel Homo
Male Female Figure 4 The use of swear words denoting women with regard to the gender of the author Figure 5 The use of swear words denoting men with regard to the gender of the author ity of occurrences of all three items are by men. The Swiss German word Schwuchtle, 'swish' is even used exclusively by male writers in Swiss WhatsApp messages. In the present pilot study, the gender of the addressee has not been taken into account, as the intendedly harmful use of gender-specific swear words only concerns non-participants of the discourse in our data. Two recently completed BA theses on a related topic include this factor in their analysis (Strebel 2019;Ramisberger 2020) and it will also be considered thoroughly in the follow-up project.

Discussion
The spectrum of the results ranges from 'rather expected' to 'very surprising' compared with the findings from previous studies. As a first observation, we found that the investigated items were never targeted offensively towards the interlocutor. This means that gendered hate speech in the context of Swiss WhatsApp messages seems to be some sort of 'indirect' hate speech, since in our data none of the examples referred to the addressee in a clearly disrespectful or intendedly harmful way. For this reason, we did not consider the gender of the addressee as a factor for our pilot study. As we have seen above, most occurrences of terms referring to the concept of 'prostitute' are used in a disrespectful and offensive context, whereas the terms fr. poule and it. fica/figa constitute a clear exception. Fr. poule is frequently used in a hypocoristic and endearing context. This is due to the fact that this word has undergone some semantic changes: etymologically, it refers to a female animal (hen). Today, it is polysemic: apart from the female animal, it can refer to a 'woman of easy virtue', a 'man's mistress', with a rather pejorative connotation (prostitute), and to the concept of '(girl) friend' (especially in expressions with a possessive as fr. ma poule). 6 As our data show, the latter constitutes the majority of occurrences of this term. The Italian term fica/figa represents a similar case. The original meaning referred to the female sex organ (Battisti, Alessio 1975), but it has also undergone some semantic changes and today it is often used to refer to an attractive woman of striking appearance (Battaglia, Barberi Squarotti 1961-2004. Accordingly, and in parallel to poule, it is very rarely used offensively in our data.
Furthermore, only 56% of the occurrences of the German word referring to the female sex organ (Futz/Fotze, 'cunt') are used disrespectfully and, more surprisingly, almost equally often by women (47%) as by men (53%). This goes against the claims put forward in Fine and Johnson (1984) and Stapleton (2003), according to which women tend to eschew swear words referring to the female sexual anatomy as they find them indecent. Moreover, only a small number of the terms referring to male homosexuals were used with an offensive meaning and when used thus, mainly by men (especially the word Schwuchtel, 'swish', which was used exclusively by male writers). It could be, though, that these two examples are instances of so-called 'reappropriation', a kind of self-labelling by pejorative terms often used to "weaken [its] stigmatizing force" (Galinsky et al. 2013(Galinsky et al. , 2020. Concerning the other swear words referring to males, our data showed that apparently, the only term regularly used with a pejorative meaning is the German word Wichser, 'wanker' (73% of the occurrences). It seems that, at least in our data, this term is as offensive for men as the terms denoting the concept 'prostitute' for women.
With regard to the use of swearwords by male and female writers in our data, we identified the word Bitch in the German corpus as an 'outlier', seeing that it is the only term used more often by women than by men. The reason for this exception is not clear, but it could have something to do with the linguistic origin of the term: Bitch is a foreign word, borrowed from English, which could mean that using swear words which do not belong to one's own mother tongue is not considered as rude as using equivalent terms from the native language. This hypothesis is confirmed by Sulpizio et al.'s study (2019) on the neural processing of so-called "taboo words" in native and foreign languages: according to the participants' responses, swear words are considered less offensive when they stem from a foreign language.
After this brief discussion of our most significant results, it should be pointed out that the present analysis is merely a lexicological pilot study and that it is by no means intended to be an exhaustive study on how Swiss WhatsApp users behave in terms of gendered hate speech. The examples we found are not sufficient in type or token numbers to draw clear conclusions, but they rather serve as a starting point for further research. There are also several aspects related to this topic that have not been discussed in this paper, but which have to be taken into account systematically in future work, first of all the gender of the messages' addressees.
Finally, the small number of gender-specific swear words used disrespectfully in our data is maybe due to the fact that in our What-sApp messages, we deal with somewhat biased data since the informants volunteered to donate their chats to the research project What's up, Switzerland?

Conclusion
Summing up, for the time being, we have relatively few examples of gendered hate speech in a huge corpus of WhatsApp chats of over 5 million tokens. Only approximately 500 occurrences of the 14 lexical items we chose to analyse in three Swiss WhatsApp sub-corpora ((Swiss) German, French and Italian) referred to a person, and hate speech seems to solely concern non-participants of the discourse, as the items used disrespectfully only referred to a third person. We identified a differentiated use of these lexical items: some, especially the polysemous ones, tend to be used almost exclusively in an ironic or endearing way (cf. fr. poule, it. figa/fica), whereas others are mostly used in the context of (indirect) hate speech. The role of semantic change and polysemy seems to be the triggering factor here and needs to be analysed more systematically. Moreover, the data show that while men use the investigated items more often, women also use them on a regular basis. The use of the term Bitch in the German sub-corpus presents a rather surprising result: it is the only gender-specific swear word that is used more frequently by women than by men. Terms referring to male homosexuals do not seem to be instances of hate speech (anymore) and are often used in an ironic or even hypocoristic manner. On the whole, the insights of our analysis show that our data provide some interesting aspects for further research on gendered hate speech in the context of Swiss WhatsApp messages. An upcoming citizen science project in collaboration with the Citizen Science Centre of the University of Zurich and ETH Zurich will provide a systematic analysis of the whole WUS corpus.