Hugging Face's logo

Datasets:
WINGNUS
/
ACL-OCL

Tasks:
Token Classification
Languages:
English
Size:
10K<n<100K
Tags:
research papers
acl
License:
Dataset card Files Community
5
ACL-OCL / Base_JSON /prefixW /json /woah /2022.woah-1.21.json
Benjamin Aw
Add updated pkl file v3
6fa4bc9
raw
history blame contribute delete
130 kB
 {
"paper_id": "2022",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T05:10:25.283781Z"
},
"title": "Users Hate Blondes: Detecting Sexism in User Comments on Online Romanian News",
"authors": [
{
"first": "Andreea-Codrina",
"middle": [],
"last": "Moldovan",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Bucharest",
"location": {
"country": "Romania"
}
},
"email": "[email protected]"
},
{
"first": "Karla",
"middle": [],
"last": "Cs\u00fcr\u00f6s",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "West University of Timis",
"location": {
"settlement": "oara",
"country": "Romania"
}
},
"email": "[email protected]"
},
{
"first": "Ana-Maria",
"middle": [],
"last": "Bucur",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Bucharest",
"location": {
"country": "Romania"
}
},
"email": "[email protected]"
},
{
"first": "Loredana",
"middle": [],
"last": "Bercuci",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "West University of Timis",
"location": {
"settlement": "oara",
"country": "Romania"
}
},
"email": "[email protected]"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "Romania ranks almost last in Europe when it comes to gender equality in political representation, with about 10% fewer women in politics than the E.U. average. We proceed from the assumption that this underrepresentation is also influenced by the sexism and verbal abuse female politicians face in the public sphere, especially in online media. We propose a novel dataset with sexist comments in Romanian language from online newspaper articles about Romanian female politicians and experiment with baseline models using classical machine learning models and fine-tuned pre-trained transformer models for the classification of sexist language in the online medium.",
"pdf_parse": {
"paper_id": "2022",
"_pdf_hash": "",
"abstract": [
{
"text": "Romania ranks almost last in Europe when it comes to gender equality in political representation, with about 10% fewer women in politics than the E.U. average. We proceed from the assumption that this underrepresentation is also influenced by the sexism and verbal abuse female politicians face in the public sphere, especially in online media. We propose a novel dataset with sexist comments in Romanian language from online newspaper articles about Romanian female politicians and experiment with baseline models using classical machine learning models and fine-tuned pre-trained transformer models for the classification of sexist language in the online medium.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "While considerable progress has been made to combat sexism in the domain of political power, we are still a long way from achieving gender balance, especially in Eastern European countries like Romania. According to the Gender Equality Index published by the European Institute for Gender Equality in 2021, in Romania, only 25.8% of ministers, 19.7% of members of parliament, and 18.4% of the members of general assemblies are women (Barbieri et al., 2021) . Romania ranks third to last (above Hungary and Malta by a small margin) regarding gender balance in political representation. It is much lower than the European Union's average (30.7% ministers, 31.5% members of parliament, and 29.3% members of general assemblies). Furthermore, women in leadership positions face discrimination and social pressure to conform to gender roles.",
"cite_spans": [
{
"start": 433,
"end": 456,
"text": "(Barbieri et al., 2021)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In the public sphere, female politicians in Romania face verbal abuse even from their peers. This abuse becomes even more prominent online, where WARNING: This paper contains sexist language. the so-called disinhibition effect leads to exaggerated behaviours, from increased self-disclosure to increased verbal violence (Joinson, 2007; Suler, 2004; Wright, 2014; Wright et al., 2018; Rodr\u00edguez-S\u00e1nchez et al., 2020) . Because the online environment offers anonymity and therefore lack of consequences for violent language, a variety of sexual slurs and sexist language appear online. This is especially true in the comment section of popular online newspapers when the topic of the articles are prominent women. The present paper uses these comments to build a dataset of sexist and non-sexist texts and experiment with several baselines models to detect the sexist language in Romanian automatically. The automatic detection of sexism could aid efforts to filter sexist texts, to encourage the prevention of, sensitization to, and sanctioning of such language.",
"cite_spans": [
{
"start": 320,
"end": 335,
"text": "(Joinson, 2007;",
"ref_id": "BIBREF20"
},
{
"start": 336,
"end": 348,
"text": "Suler, 2004;",
"ref_id": "BIBREF33"
},
{
"start": 349,
"end": 362,
"text": "Wright, 2014;",
"ref_id": "BIBREF36"
},
{
"start": 363,
"end": 383,
"text": "Wright et al., 2018;",
"ref_id": "BIBREF37"
},
{
"start": 384,
"end": 415,
"text": "Rodr\u00edguez-S\u00e1nchez et al., 2020)",
"ref_id": "BIBREF30"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Automated methods for sexism detection have been implemented using a wide range of approaches for several languages such as English Rodr\u00edguez-S\u00e1nchez et al. (2020) , Chinese (Jiang et al., 2022) , Spanish (Rodr\u00edguez-S\u00e1nchez et al., 2021) , French (Chiril et al., 2020a) or Arabic (Zahir et al., 2020) . To the best of our knowledge, there are no such studies involving Romanian. However, one Romanian dataset for offensive language has been released by Manolescu and \u00c7\u00f6ltekin (2021) . Recent studies on sexist language have been included in a review conducted by Istaiteh et al. (2020) , which also includes a literature review of racist language. Rodr\u00edguez-S\u00e1nchez et al. (2020) explore the more or less subtle ways in which sexist language manifests on Twitter in English and Spanish. The authors use a series of classical machine learning algorithms (Logistic Regression, Support Vector Machine and Random Forest) along with a Bidirectional Long Short-Term Memory model. Jha and Mamidi (2017) discuss the ambivalence of sexism and how it can be both hostile and benevolent, and achieve the best results using a FastText classifier for distinguishing between the different kinds of sexism.",
"cite_spans": [
{
"start": 132,
"end": 163,
"text": "Rodr\u00edguez-S\u00e1nchez et al. (2020)",
"ref_id": "BIBREF30"
},
{
"start": 174,
"end": 194,
"text": "(Jiang et al., 2022)",
"ref_id": "BIBREF19"
},
{
"start": 205,
"end": 237,
"text": "(Rodr\u00edguez-S\u00e1nchez et al., 2021)",
"ref_id": "BIBREF29"
},
{
"start": 247,
"end": 269,
"text": "(Chiril et al., 2020a)",
"ref_id": "BIBREF2"
},
{
"start": 280,
"end": 300,
"text": "(Zahir et al., 2020)",
"ref_id": "BIBREF38"
},
{
"start": 453,
"end": 482,
"text": "Manolescu and \u00c7\u00f6ltekin (2021)",
"ref_id": "BIBREF24"
},
{
"start": 563,
"end": 585,
"text": "Istaiteh et al. (2020)",
"ref_id": "BIBREF17"
},
{
"start": 648,
"end": 679,
"text": "Rodr\u00edguez-S\u00e1nchez et al. (2020)",
"ref_id": "BIBREF30"
},
{
"start": 974,
"end": 995,
"text": "Jha and Mamidi (2017)",
"ref_id": "BIBREF18"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "There are other similar approaches analysing how sexism manifests in different environments: at the workplace (Grosz and C\u00e9spedes, 2020) , in the gaming communities (Ghosh, 2021) , in politics (Gorrell et al., 2020; Fuchs and Sch\u00e4fer, 2020) . Several shared tasks are dedicated to the identification of online sexism and misogyny: sEXism Identification in Social neTworks (EXIST) (Rodr\u00edguez-S\u00e1nchez et al., 2021) , Automatic Misogyny Identification at IberEval (Fersini et al., 2018) and Evalita (Fersini et al., 2020) and Multimedia Automatic Misogyny Identification (MAMI) (Fersini et al., 2022) .",
"cite_spans": [
{
"start": 110,
"end": 136,
"text": "(Grosz and C\u00e9spedes, 2020)",
"ref_id": "BIBREF15"
},
{
"start": 165,
"end": 178,
"text": "(Ghosh, 2021)",
"ref_id": "BIBREF13"
},
{
"start": 193,
"end": 215,
"text": "(Gorrell et al., 2020;",
"ref_id": "BIBREF14"
},
{
"start": 216,
"end": 240,
"text": "Fuchs and Sch\u00e4fer, 2020)",
"ref_id": "BIBREF11"
},
{
"start": 380,
"end": 412,
"text": "(Rodr\u00edguez-S\u00e1nchez et al., 2021)",
"ref_id": "BIBREF29"
},
{
"start": 461,
"end": 483,
"text": "(Fersini et al., 2018)",
"ref_id": "BIBREF9"
},
{
"start": 496,
"end": 518,
"text": "(Fersini et al., 2020)",
"ref_id": "BIBREF8"
},
{
"start": 575,
"end": 597,
"text": "(Fersini et al., 2022)",
"ref_id": "BIBREF7"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "This paper proposes a novel dataset for sexist language identification in Romanian language by collecting and annotating a corpus of online comments from newspaper articles about Romanian female politicians. Furthermore, we perform experiments on this corpus using classical machine learning models and fine-tuned pre-trained transformer models, thus providing baselines for comparison in future work.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "This data used in this study is part of the ROFEM-POL Corpus 1 , which was compiled from online Romanian-language comments about Romanian female politicians. Ten of the most prominent female Romanian politicians were chosen: Clotilde Armand (USR 2 ), Viorica D\u0203ncil\u0203 (PSD 3 ), Gabriela Firea (PSD), Monica Anisie (PNL 4 ), Maria Grapini (PSD), Elena Udrea (PDL 5 ), Diana S , os , oac\u0203 (AUR 6 ), Carmen Dan (PSD), Lia Olgut , a Vasilescu (PSD). We have also included the prosecutor Laura Codrut , a Kovesi (DNA 7 / EPPO 8 ) as, although she is not technically a politician, she was an influential figure in the Romanian political scene.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "ROFEMPOL Corpus",
"sec_num": "3"
},
{
"text": "1 the corpus will be made available upon request by contacting the authors of this paper ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "ROFEMPOL Corpus",
"sec_num": "3"
},
{
"text": "A sample of 2022 comments about the aforementioned female politicians was extracted from online comments sections. All comments were available online, either from the comments section of online newspapers or Facebook pages. We have selected the most popular and active Romanian online newspapers and extracted the comments from the public news articles. The comments from Facebook were selected from the public profiles of female politicians and the online newspapers' public pages. The comments were manually extracted between November 2020 and January 2021 and reflect the salient political issues of the time. Two raters annotated the corpus, both female researchers in the field of Gender Studies, who evaluated each comment as sexist (1) or non-sexist (0).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Collection",
"sec_num": "3.1"
},
{
"text": "There are many studies that classify sexist speech and offer annotation criteria in English (Frenda et al., 2019; Parikh et al., 2019; Southern and Harmer, 2019) , French (Chiril et al., 2020b) and Indian languages (Bhattacharya et al., 2020) . Handrabura and Gherasim (2018) wrote one of the most in-depth practical guides for non-sexist speech in Romanian, classifying some of the most common types of sexist language in Romanian. As such, our annotation criteria for sexist/non-sexist texts are partially based on these studies and include: a) Gendered Violence; b) Gendered Insults; c) Role Stereotypes; d) Female Titles; e) Sexualisation and Physical Appearance.",
"cite_spans": [
{
"start": 92,
"end": 113,
"text": "(Frenda et al., 2019;",
"ref_id": "BIBREF10"
},
{
"start": 114,
"end": 134,
"text": "Parikh et al., 2019;",
"ref_id": "BIBREF25"
},
{
"start": 135,
"end": 161,
"text": "Southern and Harmer, 2019)",
"ref_id": "BIBREF31"
},
{
"start": 171,
"end": 193,
"text": "(Chiril et al., 2020b)",
"ref_id": "BIBREF3"
},
{
"start": 215,
"end": 242,
"text": "(Bhattacharya et al., 2020)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Annotation Criteria",
"sec_num": "3.2"
},
{
"text": "By gendered violence, we mean language-based sexual harassment, inciting sexual harassment and threats of physical abuse, rape, or murder. Our understanding of gendered violence is based on the concept of \"cyberviolence against women and girls\" (cyber VAWG), which is defined as the \"full spectrum of behaviour ranges from online harassment to the desire to inflict physical harm including sexual assaults, murders and suicides\" 9 . The 2015 report includes harassment in its categorization of cyber VAWG, defining it as \"the use of technology to continuously contact, annoy, threaten, and/or scare the victim\". While online comments are not sent directly to the victims, they act as defamatory language, which affects the target. We, therefore, considered these comments as harassment.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "a) Gendered Violence",
"sec_num": null
},
{
"text": "Online harassment often contains gendered insults, related especially to slut-shaming or shaming for nonconformity to societal expectations. Gendered insults represent any words or phrases used disproportionately against a specific gender and are linked to the perpetuation of stereotypical societal beliefs about that particular gender, in this case, women. Gendered insults can be i) Sexual or ii) Non-sexual. ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "b) Gendered Insults",
"sec_num": null
},
{
"text": "i) Sexual",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "b) Gendered Insults",
"sec_num": null
},
{
"text": "The gendered insults in this category denote the users' beliefs that female politicians are sexually promiscuous and have engaged in sexual favors to advance their political careers. Examples include: \"prostituat\u0203\" (EN: \"prostitute\"), \"fuf\u0203\" (EN: \"philanderer\"), \"curv\u0203\" (EN: \"whore\"), \"BITCH\", \"tarf\u0203\" (EN: \"slut\"), \"matracuc\u0203\" (EN: \"lower class and unintelligent woman\"), \"zdreant ,\u0203 \" (EN: \"disgraced woman\" and \"rag\"). We present some examples in Table 1 .",
"cite_spans": [],
"ref_spans": [
{
"start": 451,
"end": 458,
"text": "Table 1",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "b) Gendered Insults",
"sec_num": null
},
{
"text": "In this category, we have aggregated insults that do not discuss the politicians' sexual lives, but their personalities, attitudes and social statuses. Notably, in Romanian, the following examples are solely used with reference to women: \"tut\u0203\" (EN: \"dumbass\"), \"t , at ,\u0203 \" (EN: \"vulgar woman\"), \"mahalagioac\u0203\" (EN: \"loud lower-class woman who lives in the ghetto\"), \"div\u0203\", \"t , oap\u0203\" (EN: \"boor\").",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "ii) Non-sexual",
"sec_num": null
},
{
"text": "Related to the issue of gendered insults is that For our corpus, we divided female stereotypes into i) \"Women's jobs\" and ii) \"Female attitudes.\" i) \"Women's Jobs\" \"Women's jobs\" can be defined as being related to domesticity, namely being a wife, a mother, or a housekeeper. Despite their accomplishments, female politicians are oftentimes ridiculed for not having children. Users also make attempts to dismiss female politicians by claiming that they are impostors who, in fact, have stereotypically female careers (e.g., cook, washerwoman, cleaning lady, secretary) (Table 2 ).",
"cite_spans": [],
"ref_spans": [
{
"start": 569,
"end": 577,
"text": "(Table 2",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "c) Role Stereotypes",
"sec_num": null
},
{
"text": "ii) \"Female Attitudes\"",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "c) Role Stereotypes",
"sec_num": null
},
{
"text": "On the one hand, multiple comments employ positive stereotypes to talk about female politicians' success, attributing it to them being caring, maternal, soft or submissive. However, on the other hand, we also have examples of negative stereotypes, such as women being portrayed as hysterical (\"isteric\u0203\"), angry (\"nervoas\u0203\", \"crizat\u0203\"), gossipy (\"b\u00e2rfitoare\") as seen in Table 3 . Users generally employ mocking female titles when addressing female politicians. Examples include: \"cucoan\u0203\" (EN: lady), \"tanti\" (EN: auntie, referring to an older woman), \"madam\u0203\" (adapted from the French madame), \"duamn\u0203\" from \"doamn\u0203\" (EN: Mrs.), \"domnis , oar\u0203\" (EN: Miss) (Table 4) . Young or young-looking politicians are also regularly addressed as \"fat\u0203\" or \"fetis , can\u0203\" (EN: girl or girlie). There are also cases of users creating female versions of male titles with negative connotations: e.g., for the female minister, \"ministru\" (EN: minister) is mixed with \"monstruas\u0203\" (EN: monstrous) to create \"ministruoas\u0203\" or with \"stripteuz\u0203\" (EN: stripper) to create \"min-istreuz\u0203\". Another example is \"prim\u0203rit ,\u0203 \" (EN: female mayor), which adds the diminutive \"-it ,\u0203 \" to \"primar\" (EN: mayor). Such forms of address are used ironically in an attempt to deride the politicians and to express contempt.",
"cite_spans": [],
"ref_spans": [
{
"start": 371,
"end": 378,
"text": "Table 3",
"ref_id": "TABREF3"
},
{
"start": 658,
"end": 667,
"text": "(Table 4)",
"ref_id": "TABREF4"
}
],
"eq_spans": [],
"section": "c) Role Stereotypes",
"sec_num": null
},
{
"text": "Comments on the physical appearance of the female politicians were very frequent, even though they were generally irrelevant to the topics of the news stories. Physical appearance comments include comments on: i) Body Weight, ii) Hair Color Stereotypes, iii) Clothes and Style, and iv) Perceived Physical Attractiveness.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "e) Sexualisation and Physical Appearance",
"sec_num": null
},
{
"text": "Body shaming comments were more frequent in the case of female politicians who do not conform to the \"thin\" beauty standard. In this case, politicians were directly called \"gras\u0203\" (EN: \"fat\"), or were compared to animals and fantastical creatures that are stereotyped as being overweight, e.g., \"porc\" (EN: \"pig\") or \"matahal\u0203\" (EN: \"bugbear\") as seen in Table 5 . ii) Hair Color Stereotypes The most common stereotype about hair color in the corpus is the \"dumb blonde\". As such, blonde female politicians received a significant number of comments in which users would make derogatory references to their hair color, often identifying them with it (Table 6 ).",
"cite_spans": [],
"ref_spans": [
{
"start": 355,
"end": 362,
"text": "Table 5",
"ref_id": "TABREF5"
},
{
"start": 649,
"end": 657,
"text": "(Table 6",
"ref_id": "TABREF6"
}
],
"eq_spans": [],
"section": "i) Body Weight",
"sec_num": null
},
{
"text": "News stories that feature full-body pictures of female politicians often drew comments about their fashion style and choices. For example, references were made to the cleanliness of their clothes, the shortness of their skirts (\"fust\u0203 scurt\u0203\"), as well as the expensiveness of their outfits (i.e., clothes and accessories), as seen in Table 7 . iv) Perceived Physical Attractiveness Commenters offer judgements on how attractive or unattractive they find certain female politi-cians. Both negative -e.g. \"ur\u00e2t\u0203\" (EN: \"ugly\"), \"hidos , enie\" (EN: \"monstrosity\") -as well as positive -e.g. \"frumoas\u0203\" (EN: \"beautiful\") -were included as being sexist, as they are both inappropriate in context. Moreover, for the \"most unattractive\" politicians, commenters questioned their female gender, claiming that they are not women (\"nu e femeie\") as seen in Table 8 .",
"cite_spans": [],
"ref_spans": [
{
"start": 335,
"end": 342,
"text": "Table 7",
"ref_id": "TABREF7"
},
{
"start": 846,
"end": 853,
"text": "Table 8",
"ref_id": "TABREF8"
}
],
"eq_spans": [],
"section": "iii) Clothes and Style",
"sec_num": null
},
{
"text": "Dehumanisation occurs primarily through zoomorphism, i.e., by comparing the targets with female animals, often with sexual connotations. Some examples (Table 9) include \"creatur\u0203\" (EN: \"creature\"), \"vac\u0203\" (EN: \"cow\"), \"javr\u0203\" (EN: \"female dog\", \"bitch\"), \"scroaf\u0203\" (EN: \"sow\").",
"cite_spans": [],
"ref_spans": [
{
"start": 151,
"end": 160,
"text": "(Table 9)",
"ref_id": "TABREF9"
}
],
"eq_spans": [],
"section": "v) Dehumanisation",
"sec_num": null
},
{
"text": "After both raters annotated the data individually, we measured the inter-rater agreement using Cohen's Kappa. The coefficient can have values between -1 and 1, with a value equal to 1 meaning a perfect agreement between the annotators. The Cohen's Kappa coefficient for the annotations of the two raters in this paper is 0.87, meaning there was a good inter-rater agreement between the annotators. The two raters discussed the disagreements until a final common decision was made. The resulting ROFEMPOL dataset contains 1135 non-sexist samples and 887 sexist samples.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "v) Dehumanisation",
"sec_num": null
},
{
"text": "We explored the dataset by computing the keyness scores for the sexist and non-sexist texts (Kilgarriff, 2009; Gabrielatos, 2018) . The keyness analysis is performed by comparing the frequencies of the words from the sexist comments (target corpus) to the frequencies of words from the non-sexist comments (reference corpus). In Figure 1 we report the top 15 words from the two classes ordered by their log-likelihood ratio (G 2 ) (Dunning, 1993).",
"cite_spans": [
{
"start": 92,
"end": 110,
"text": "(Kilgarriff, 2009;",
"ref_id": "BIBREF21"
},
{
"start": 111,
"end": 129,
"text": "Gabrielatos, 2018)",
"ref_id": "BIBREF12"
}
],
"ref_spans": [
{
"start": 329,
"end": 337,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "v) Dehumanisation",
"sec_num": null
},
{
"text": "The sexist comments contain some sexist keywords such as \"madam\" (EN: \"madam\"), \"madame\" (EN: \"madam\"), \"mahalagioaic\u0203\" (EN: \"ghetto woman\") or \"coan\u0203\" (EN: \"lady\"), while non-sexist texts contain more polite addressing forms such as \"doamn\u0203\" (EN: \"Mrs.\") or \"dumneavoastr\u0203\" (EN: \"you\", the polite second person singular or plural in Romanian language).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "v) Dehumanisation",
"sec_num": null
},
{
"text": "In this section, we propose several baseline methods for the binary classification of sexist language using ROFEMPOL. We explore several encoding methods for the Romanian text data, such as Bagof-Words and BERT-based sentence representa- tions. Alongside classical machine learning models (i.e., Logistic Regression, SVM, Random Forest), we also explore fine-tuning pre-trained transformer models for the Romanian language.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Baseline Methods",
"sec_num": "4"
},
{
"text": "Bag-of-Words (BOW) and Term Frequency-Inverse Document Frequency (TF-IDF) We chose to use BOW-based representations because they are language independent. In addition, BOW representations allow for modelling sexism based on keywords, as some sentences in the dataset can be easily identified based on certain sexist words (i.e., the keywords from Figure 1) .",
"cite_spans": [],
"ref_spans": [
{
"start": 347,
"end": 356,
"text": "Figure 1)",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Text Representation",
"sec_num": "4.1"
},
{
"text": "1 sarcastic adaptation of the French madame; the word suggests the person does not deserve the title of lady, madam 2 a loud lower class woman who is unrefined 3 slang specifically used to address women directly, similar to lady, but it implies the woman is lower class 4 it suggests that the woman addressed is older, unattractive and unrefined 5 of one of the politicians 6 the polite second person singular or plural 7 the polite second person singular or plural, or plain form of second person plural Multilingual BERT As opposed to the BOW and TF-IDF word representations, which do not contain any information about the context, sentence representations as given by modern transformer networks (Reimers and Gurevych, 2019) offer richer semantic information and have been successfully used in low-resource scenarios (Ranasinghe and Zampieri, 2021) . As such, we use Sentence Transformer (Reimers and Gurevych, 2019) to extract embeddings from BERT-based models. We use a pre-trained Multilingual BERT (M-BERT) (Devlin et al., 2019) which was trained on 102 languages using Wikipedia text, including Romanian language.",
"cite_spans": [
{
"start": 699,
"end": 727,
"text": "(Reimers and Gurevych, 2019)",
"ref_id": "BIBREF28"
},
{
"start": 820,
"end": 851,
"text": "(Ranasinghe and Zampieri, 2021)",
"ref_id": "BIBREF27"
},
{
"start": 891,
"end": 919,
"text": "(Reimers and Gurevych, 2019)",
"ref_id": "BIBREF28"
},
{
"start": 1014,
"end": 1035,
"text": "(Devlin et al., 2019)",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Text Representation",
"sec_num": "4.1"
},
{
"text": "Romanian BERT As opposed to M-BERT, the Romanian BERT (Ro-BERT) (Dumitrescu et al., 2020) , is a more specialized model, trained on a larger Romanian corpus. Moreover, the tokenizer is better suited for handling Romanian texts, using fewer tokens to encode words than M-BERT, while also having fewer unknown tokens. The model was trained on a large corpus of Romanian data from Wikipedia, OPUS (parallel corpus with translated texts from the web) (Tiedemann, 2012) and OSCAR (Common Crawl data in Romanian language) (Su\u00e1rez et al., 2019) . We use Ro-BERT for extracting sentence representations.",
"cite_spans": [
{
"start": 64,
"end": 89,
"text": "(Dumitrescu et al., 2020)",
"ref_id": "BIBREF5"
},
{
"start": 447,
"end": 464,
"text": "(Tiedemann, 2012)",
"ref_id": "BIBREF35"
},
{
"start": 516,
"end": 537,
"text": "(Su\u00e1rez et al., 2019)",
"ref_id": "BIBREF32"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Text Representation",
"sec_num": "4.1"
},
{
"text": "We evaluate the performance of classical machine learning classifiers: Logistic Regression (LR), Random Forest (RF) and Support Vector Machine (SVM) on BOW and semantic sentence representations. Moreover, we directly fine-tune transformer models pre-trained on Romanian language text: M-BERT and Ro-BERT. The Ro-BERT model has been shown to outperform its multilingual counterpart, M-BERT, in downstream tasks such as named entity recognition and part-of-speech tagging (Dumitrescu et al., 2020).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "4.2"
},
{
"text": "In this section, we describe the classification experiments performed for the detection of sexist language in Romanian text and report the obtained results.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments and Results",
"sec_num": "5"
},
{
"text": "ROFEMPOL Corpus is split into training and testing sets, with 1617 texts in the training split and 405 texts in the testing split.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments",
"sec_num": "5.1"
},
{
"text": "Since the dataset contains a small number of Table 10 : Results for sexist language detection on ROFEMPOL. We report Precision, Recall and F 1 for each model on the two classes (Non-sexist and Sexist) and weighted averages. We also report Macro-F 1 score. The best performing model is the fine-tuned Ro-BERT. samples, we performed a 5-fold cross-validation for all models. For LR, RF and SVM, we perform a hyperparameter grid search on each fold to find the best hyperparameters for the models. The search space used for grid search for each model can be found in Appendix.",
"cite_spans": [],
"ref_spans": [
{
"start": 45,
"end": 53,
"text": "Table 10",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Experiments",
"sec_num": "5.1"
},
{
"text": "The pre-trained transformer models, Ro-BERT and M-BERT, are fine-tuned using the AdamW (Kingma and Ba, 2014) optimizer with a learning rate of 0.00001 with a linear decay with 50 warmup steps. Due to computational limitations, we train the models for 4 epochs with a batch size of 4.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments",
"sec_num": "5.1"
},
{
"text": "All the reported results are obtained from the 5-fold cross-validation. The performance of models from each fold is evaluated on the test split. We report the mean and standard deviation of the performance scores for the 5 splits for Precision, Recall and F 1 -score, weighted averages and macro-F 1 .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments",
"sec_num": "5.1"
},
{
"text": "The results of the classification experiments for sexism classification on the ROFEMPOL Corpus are presented in Table 10 . The best performing model is Ro-BERT, the pre-trained BERT model for Romanian language, obtaining the overall best scores in discriminating between sexist and nonsexist texts. Ro-BERT attains a 0.75 Macro-F 1 score in the classification task, an improvement of 0.03 over M-BERT.",
"cite_spans": [],
"ref_spans": [
{
"start": 112,
"end": 120,
"text": "Table 10",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Results",
"sec_num": "5.2"
},
{
"text": "All the models perform better at classifying the non-sexist texts than identifying the sexist ones. The differences between the performances for the two classes are greater for the classical machine learning models using BOW, TF-IDF and BERT-based representations than for the fine-tunes models. The biggest gap in the performances between the two classes is found in the classification using Random Forest on Bag-of-Words representations, with 0.74 F 1 -score for non-sexist and 0.44 F 1 -score for the sexist class.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results",
"sec_num": "5.2"
},
{
"text": "Comparing the three classical machine learning models, SVM and LR perform better than RF for all the text representation methods (BOW-based, BERT-based encodings), as seen in the Macro-F 1 score.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results",
"sec_num": "5.2"
},
{
"text": "We present a sample of correct and incorrect predictions from the best performing model, Ro-BERT, in Table 11 and 12. Figure 2: Visualizing salient tokens contributing to the prediction of the correct class using the Ro-BERT model. Term color indicates attribution intensity, red is negative, green is positive. EN: \"The right man at the right place ! Good luck !\", \"Madam you are lame ...\" Figure 3 : Visualizing salient tokens contributing to the prediction of the incorrect class using the Ro-BERT model. Term color indicates attribution intensity, red is negative, green is positive. EN: \"Dumbass you have already managed to destroy a generation from Romania's future!\", \"Once upon a time there was a Diana princess of hearts, now God has sent us a Diana princess of the Romanians.\" Regarding the incorrect decisions of the model, some of the texts predicted as being sexist contain offensive words, but the words are not sexist, according to the annotations guidelines, such as \"analfabet\u0203\" (EN: \"illiterate\") or incult\u0203 (EN: \"uneducated\"). Other offensive words are not targeted towards female politicians but other people. For example, \"cretinilor\" (EN: \"idiots\") is targeting voters. The texts incorrectly labelled as non-sexist contain vulgar, sexist words such as \"t , oap\u0203\" (EN: \"boor\") and \"tut\u0203\" (EN: \"dumbass\") that can be easily recognised as being sexist. The Ro-BERT model may fail to recognise these words as sexist because it is not pre-trained on informal text found in the comments from news articles or social media.",
"cite_spans": [],
"ref_spans": [
{
"start": 101,
"end": 109,
"text": "Table 11",
"ref_id": "TABREF1"
},
{
"start": 391,
"end": 399,
"text": "Figure 3",
"ref_id": null
}
],
"eq_spans": [],
"section": "Qualitative Results and Discussion",
"sec_num": "5.3"
},
{
"text": "Furthermore, we compute word importance (attribution scores) from Ro-BERT to interpret the model's predictions. We use Integrated Gradients (Sundararajan et al., 2017) from the Captum library (Kokhlikyan et al., 2020) to show the most salient tokens for the incorrect predictions. From the examples in Figure 2 , we can conclude that the model can recognize \"madam\" being used as a sexist word in the samples from our corpus, thus having an important attribution in the decision of the model. In Figure 3 we show two examples of word attributions in incorrectly predicted texts. In the first comment, even if the tokens tut ##o (EN: \"dumbass\") have strong negative attributions, the final decision is also influenced by the other tokens in the sentence, labelling the comment as being nonsexist. In the second text, the token \"print , es\u0203\" (EN: \"princess\") has a strong positive attribution in the decision of the model to label the text as being sexist, although the word is not used as sexist in this context.",
"cite_spans": [
{
"start": 140,
"end": 167,
"text": "(Sundararajan et al., 2017)",
"ref_id": "BIBREF34"
},
{
"start": 192,
"end": 217,
"text": "(Kokhlikyan et al., 2020)",
"ref_id": "BIBREF23"
}
],
"ref_spans": [
{
"start": 302,
"end": 310,
"text": "Figure 2",
"ref_id": null
},
{
"start": 496,
"end": 504,
"text": "Figure 3",
"ref_id": null
}
],
"eq_spans": [],
"section": "Qualitative Results and Discussion",
"sec_num": "5.3"
},
{
"text": "Our findings underline the fact that Romanian female politicians are relentlessly targeted and stereotyped because of their gender in the eye of the Romanian public, who uses sexist language to criticise them online. Automatic detection of sexist speech in Romanian language is a result of the imperative need to eliminate such forms of gender-based discrimination in order to work towards gender equality and a truly democratic Romanian society, one in which female politicians can thrive. We presented the novel ROFEMPOL dataset for sexist language identification in Romanian, collected from online comments from newspaper articles about Romanian female politicians. Furthermore, we performed experiments on this corpus using classical machine learning models and fine-tuned pre-trained transformer models, thus providing baselines for comparison in future work.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "6"
},
{
"text": "Further work on the ROFEMPOL corpus will attempt to include a larger dataset and to annotate the categories outlined in this paper. The ROFEMPOL dataset could also prove productive in a comparison with a corpus on male politicians, testing whether sexist speech is a general phenomenon for Romanian users or whether it is only targeted at female politicians. Lastly, ROFEMPOL could be included in a larger corpus on sexism in the Romanian language in an attempt to limit the spread of sexist language online.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "6"
},
{
"text": "https://www.broadbandcommission.org/publication/cyberviolence-against-women/ (last accessed April 10, 2022)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Julia Ochmann, and Vytautas Peciukonis. 2021. Gender equality index 2021",
"authors": [
{
"first": "Davide",
"middle": [],
"last": "Barbieri",
"suffix": ""
},
{
"first": "Antonio",
"middle": [],
"last": "Garcia Cazorla",
"suffix": ""
},
{
"first": "Thil",
"middle": [],
"last": "Laur\u00e8ne",
"suffix": ""
},
{
"first": "Blandine",
"middle": [],
"last": "Mollard",
"suffix": ""
}
],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Davide Barbieri, Antonio Garcia Cazorla, Laur\u00e8ne THIL, Blandine Mollard, Julia Ochmann, and Vy- tautas Peciukonis. 2021. Gender equality index 2021: Health.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Akash Bhagat, Yogesh Dawer, Bornini Lahiri, and Atul Kr. Ojha. 2020. Developing a multilingual annotated corpus of misogyny and aggression",
"authors": [
{
"first": "Shiladitya",
"middle": [],
"last": "Bhattacharya",
"suffix": ""
},
{
"first": "Siddharth",
"middle": [],
"last": "Singh",
"suffix": ""
},
{
"first": "Ritesh",
"middle": [],
"last": "Kumar",
"suffix": ""
},
{
"first": "Akanksha",
"middle": [],
"last": "Bansal",
"suffix": ""
}
],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.48550/ARXIV.2003.07428"
]
},
"num": null,
"urls": [],
"raw_text": "Shiladitya Bhattacharya, Siddharth Singh, Ritesh Ku- mar, Akanksha Bansal, Akash Bhagat, Yogesh Dawer, Bornini Lahiri, and Atul Kr. Ojha. 2020. De- veloping a multilingual annotated corpus of misogyny and aggression.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "An annotated corpus for sexism detection in french tweets",
"authors": [
{
"first": "Patricia",
"middle": [],
"last": "Chiril",
"suffix": ""
},
{
"first": "V\u00e9ronique",
"middle": [],
"last": "Moriceau",
"suffix": ""
},
{
"first": "Farah",
"middle": [],
"last": "Benamara",
"suffix": ""
},
{
"first": "Alda",
"middle": [],
"last": "Mari",
"suffix": ""
},
{
"first": "Gloria",
"middle": [],
"last": "Origgi",
"suffix": ""
},
{
"first": "Marl\u00e8ne",
"middle": [],
"last": "Coulomb-Gully",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 12th language resources and evaluation conference",
"volume": "",
"issue": "",
"pages": "1397--1403",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Patricia Chiril, V\u00e9ronique Moriceau, Farah Benamara, Alda Mari, Gloria Origgi, and Marl\u00e8ne Coulomb- Gully. 2020a. An annotated corpus for sexism de- tection in french tweets. In Proceedings of the 12th language resources and evaluation conference, pages 1397-1403.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "An annotated corpus for sexism detection in French tweets",
"authors": [
{
"first": "Patricia",
"middle": [],
"last": "Chiril",
"suffix": ""
},
{
"first": "V\u00e9ronique",
"middle": [],
"last": "Moriceau",
"suffix": ""
},
{
"first": "Farah",
"middle": [],
"last": "Benamara",
"suffix": ""
},
{
"first": "Alda",
"middle": [],
"last": "Mari",
"suffix": ""
},
{
"first": "Gloria",
"middle": [],
"last": "Origgi",
"suffix": ""
},
{
"first": "Marl\u00e8ne",
"middle": [],
"last": "Coulomb-Gully",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 12th Language Resources and Evaluation Conference",
"volume": "",
"issue": "",
"pages": "1397--1403",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Patricia Chiril, V\u00e9ronique Moriceau, Farah Benamara, Alda Mari, Gloria Origgi, and Marl\u00e8ne Coulomb- Gully. 2020b. An annotated corpus for sexism detec- tion in French tweets. In Proceedings of the 12th Lan- guage Resources and Evaluation Conference, pages 1397-1403, Marseille, France. European Language Resources Association.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "BERT: Pre-training of deep bidirectional transformers for language understanding",
"authors": [
{
"first": "Jacob",
"middle": [],
"last": "Devlin",
"suffix": ""
},
{
"first": "Ming-Wei",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "Kenton",
"middle": [],
"last": "Lee",
"suffix": ""
},
{
"first": "Kristina",
"middle": [],
"last": "Toutanova",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
"volume": "1",
"issue": "",
"pages": "4171--4186",
"other_ids": {
"DOI": [
"10.18653/v1/N19-1423"
]
},
"num": null,
"urls": [],
"raw_text": "Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language under- standing. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Tech- nologies, Volume 1 (Long and Short Papers), pages 4171-4186, Minneapolis, Minnesota. Association for Computational Linguistics.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "The birth of Romanian BERT",
"authors": [
{
"first": "Stefan",
"middle": [],
"last": "Dumitrescu",
"suffix": ""
},
{
"first": "Andrei-Marius",
"middle": [],
"last": "Avram",
"suffix": ""
},
{
"first": "Sampo",
"middle": [],
"last": "Pyysalo",
"suffix": ""
}
],
"year": 2020,
"venue": "Findings of the Association for Computational Linguistics: EMNLP 2020",
"volume": "",
"issue": "",
"pages": "4324--4328",
"other_ids": {
"DOI": [
"10.18653/v1/2020.findings-emnlp.387"
]
},
"num": null,
"urls": [],
"raw_text": "Stefan Dumitrescu, Andrei-Marius Avram, and Sampo Pyysalo. 2020. The birth of Romanian BERT. In Findings of the Association for Computational Lin- guistics: EMNLP 2020, pages 4324-4328, Online. Association for Computational Linguistics.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Accurate methods for the statistics of surprise and coincidence",
"authors": [
{
"first": "E",
"middle": [],
"last": "Ted",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Dunning",
"suffix": ""
}
],
"year": 1993,
"venue": "Computational linguistics",
"volume": "19",
"issue": "1",
"pages": "61--74",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ted E Dunning. 1993. Accurate methods for the statis- tics of surprise and coincidence. Computational lin- guistics, 19(1):61-74.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "SemEval-2022 task 5: Multimedia automatic misogyny identification",
"authors": [
{
"first": "Elisabetta",
"middle": [],
"last": "Fersini",
"suffix": ""
},
{
"first": "Francesca",
"middle": [],
"last": "Gasparini",
"suffix": ""
},
{
"first": "Giulia",
"middle": [],
"last": "Rizzi",
"suffix": ""
},
{
"first": "Aurora",
"middle": [],
"last": "Saibene",
"suffix": ""
},
{
"first": "Berta",
"middle": [],
"last": "Chulvi",
"suffix": ""
},
{
"first": "Paolo",
"middle": [],
"last": "Rosso",
"suffix": ""
},
{
"first": "Alyssa",
"middle": [],
"last": "Lees",
"suffix": ""
},
{
"first": "Jeffrey",
"middle": [],
"last": "Sorensen",
"suffix": ""
}
],
"year": 2022,
"venue": "Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022). Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Elisabetta Fersini, Francesca Gasparini, Giulia Rizzi, Aurora Saibene, Berta Chulvi, Paolo Rosso, Alyssa Lees, and Jeffrey Sorensen. 2022. SemEval-2022 task 5: Multimedia automatic misogyny identifica- tion. In Proceedings of the 16th International Work- shop on Semantic Evaluation (SemEval-2022). Asso- ciation for Computational Linguistics.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Ami@ evalita2020: Automatic misogyny identification",
"authors": [
{
"first": "Elisabetta",
"middle": [],
"last": "Fersini",
"suffix": ""
},
{
"first": "Debora",
"middle": [],
"last": "Nozza",
"suffix": ""
},
{
"first": "Paolo",
"middle": [],
"last": "Rosso",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Elisabetta Fersini, Debora Nozza, and Paolo Rosso. 2020. Ami@ evalita2020: Automatic misogyny iden- tification. In EVALITA.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Overview of the task on automatic misogyny identification at ibereval 2018. Ibereval@ sepln",
"authors": [
{
"first": "Elisabetta",
"middle": [],
"last": "Fersini",
"suffix": ""
},
{
"first": "Paolo",
"middle": [],
"last": "Rosso",
"suffix": ""
},
{
"first": "Maria",
"middle": [],
"last": "Anzovino",
"suffix": ""
}
],
"year": 2018,
"venue": "",
"volume": "2150",
"issue": "",
"pages": "214--228",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Elisabetta Fersini, Paolo Rosso, and Maria Anzovino. 2018. Overview of the task on automatic misog- yny identification at ibereval 2018. Ibereval@ sepln, 2150:214-228.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Online hate speech against women: Automatic identification of misogyny and sexism on twitter",
"authors": [
{
"first": "Simona",
"middle": [],
"last": "Frenda",
"suffix": ""
},
{
"first": "Bilal",
"middle": [],
"last": "Ghanem",
"suffix": ""
},
{
"first": "Manuel",
"middle": [],
"last": "Montes",
"suffix": ""
},
{
"first": "Paolo",
"middle": [],
"last": "Rosso",
"suffix": ""
}
],
"year": 2019,
"venue": "Journal of Intelligent & Fuzzy Systems",
"volume": "36",
"issue": "",
"pages": "4743--4752",
"other_ids": {
"DOI": [
"10.3233/JIFS-179023"
]
},
"num": null,
"urls": [],
"raw_text": "Simona Frenda, Bilal Ghanem, Manuel Montes, and Paolo Rosso. 2019. Online hate speech against women: Automatic identification of misogyny and sexism on twitter. Journal of Intelligent & Fuzzy Systems, 36:4743-4752.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Normalizing misogyny: hate speech and verbal abuse of female politicians on japanese twitter",
"authors": [
{
"first": "Tamara",
"middle": [],
"last": "Fuchs",
"suffix": ""
},
{
"first": "Fabian",
"middle": [],
"last": "Sch\u00e4fer",
"suffix": ""
}
],
"year": 2020,
"venue": "Japan Forum",
"volume": "33",
"issue": "",
"pages": "1--27",
"other_ids": {
"DOI": [
"10.1080/09555803.2019.1687564"
]
},
"num": null,
"urls": [],
"raw_text": "Tamara Fuchs and Fabian Sch\u00e4fer. 2020. Normalizing misogyny: hate speech and verbal abuse of female politicians on japanese twitter. Japan Forum, 33:1- 27.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Keyness analysis: Nature, metrics and techniques",
"authors": [
{
"first": "Costas",
"middle": [],
"last": "Gabrielatos",
"suffix": ""
}
],
"year": 2018,
"venue": "Corpus Approaches to Discourse",
"volume": "",
"issue": "",
"pages": "225--258",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Costas Gabrielatos. 2018. Keyness analysis: Nature, metrics and techniques. In Corpus Approaches to Discourse, pages 225-258. Routledge.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Analyzing toxicity in online gaming communities",
"authors": [
{
"first": "Ayushi",
"middle": [],
"last": "Ghosh",
"suffix": ""
}
],
"year": 2021,
"venue": "Turkish Journal of Computer and Mathematics Education (TURCOMAT)",
"volume": "12",
"issue": "10",
"pages": "4448--4455",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ayushi Ghosh. 2021. Analyzing toxicity in online gam- ing communities. Turkish Journal of Computer and Mathematics Education (TURCOMAT), 12(10):4448- 4455.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Which politicians receive abuse? four factors illuminated in the uk general election 2019",
"authors": [
{
"first": "Genevieve",
"middle": [],
"last": "Gorrell",
"suffix": ""
},
{
"first": "Mehmet",
"middle": [],
"last": "Bakir",
"suffix": ""
},
{
"first": "Ian",
"middle": [],
"last": "Roberts",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Greenwood",
"suffix": ""
},
{
"first": "Kalina",
"middle": [],
"last": "Bontcheva",
"suffix": ""
}
],
"year": 2020,
"venue": "EPJ Data Science",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.1140/epjds/s13688-020-00236-9"
]
},
"num": null,
"urls": [],
"raw_text": "Genevieve Gorrell, Mehmet Bakir, Ian Roberts, Mark Greenwood, and Kalina Bontcheva. 2020. Which politicians receive abuse? four factors illuminated in the uk general election 2019. EPJ Data Science, 9.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Automatic detection of sexist statements commonly used at the workplace. CoRR, abs",
"authors": [
{
"first": "Dylan",
"middle": [],
"last": "Grosz",
"suffix": ""
},
{
"first": "Patricia",
"middle": [
"Conde"
],
"last": "C\u00e9spedes",
"suffix": ""
}
],
"year": 2007,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Dylan Grosz and Patricia Conde C\u00e9spedes. 2020. Auto- matic detection of sexist statements commonly used at the workplace. CoRR, abs/2007.04181.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Limbajul nonsexist: Repere conceptuale\u015fi recomand\u0203ri practice",
"authors": [
{
"first": "L",
"middle": [],
"last": "Handrabura",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Gherasim",
"suffix": ""
}
],
"year": 2018,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "L Handrabura and A Gherasim. 2018. Limbajul nonsex- ist: Repere conceptuale\u015fi recomand\u0203ri practice.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Racist and sexist hate speech detection: Literature review",
"authors": [
{
"first": "Othman",
"middle": [],
"last": "Istaiteh",
"suffix": ""
},
{
"first": "Razan",
"middle": [],
"last": "Al-Omoush",
"suffix": ""
},
{
"first": "Sara",
"middle": [],
"last": "Tedmori",
"suffix": ""
}
],
"year": 2020,
"venue": "2020 International Conference on Intelligent Data Science Technologies and Applications (IDSTA)",
"volume": "",
"issue": "",
"pages": "95--99",
"other_ids": {
"DOI": [
"10.1109/IDSTA50958.2020.9264052"
]
},
"num": null,
"urls": [],
"raw_text": "Othman Istaiteh, Razan Al-Omoush, and Sara Tedmori. 2020. Racist and sexist hate speech detection: Liter- ature review. In 2020 International Conference on Intelligent Data Science Technologies and Applica- tions (IDSTA), pages 95-99.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "When does a compliment become sexist? analysis and classification of ambivalent sexism using twitter data",
"authors": [
{
"first": "Akshita",
"middle": [],
"last": "Jha",
"suffix": ""
},
{
"first": "Radhika",
"middle": [],
"last": "Mamidi",
"suffix": ""
}
],
"year": 2017,
"venue": "Proceedings of the Second Workshop on NLP and Computational Social Science",
"volume": "",
"issue": "",
"pages": "7--16",
"other_ids": {
"DOI": [
"10.18653/v1/W17-2902"
]
},
"num": null,
"urls": [],
"raw_text": "Akshita Jha and Radhika Mamidi. 2017. When does a compliment become sexist? analysis and classifi- cation of ambivalent sexism using twitter data. In Proceedings of the Second Workshop on NLP and Computational Social Science, pages 7-16, Vancou- ver, Canada. Association for Computational Linguis- tics.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "Swsr: A chinese dataset and lexicon for online sexism detection",
"authors": [
{
"first": "Aiqi",
"middle": [],
"last": "Jiang",
"suffix": ""
},
{
"first": "Xiaohan",
"middle": [],
"last": "Yang",
"suffix": ""
},
{
"first": "Yang",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Arkaitz",
"middle": [],
"last": "Zubiaga",
"suffix": ""
}
],
"year": 2022,
"venue": "Online Social Networks and Media",
"volume": "27",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Aiqi Jiang, Xiaohan Yang, Yang Liu, and Arkaitz Zubi- aga. 2022. Swsr: A chinese dataset and lexicon for online sexism detection. Online Social Networks and Media, 27:100182.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "Causes and implications of disinhibited behavior on the internet. Psychology and the internet: intrapersonal, interpersonal, and transpersonal implications",
"authors": [
{
"first": "Adam",
"middle": [],
"last": "Joinson",
"suffix": ""
}
],
"year": 2007,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Adam Joinson. 2007. Causes and implications of disin- hibited behavior on the internet. Psychology and the internet: intrapersonal, interpersonal, and transper- sonal implications.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Simple maths for keywords",
"authors": [
{
"first": "Adam",
"middle": [],
"last": "Kilgarriff",
"suffix": ""
}
],
"year": 2009,
"venue": "Proc. Corpus Linguistics",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Adam Kilgarriff. 2009. Simple maths for keywords. In Proc. Corpus Linguistics.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "Adam: A method for stochastic optimization",
"authors": [
{
"first": "P",
"middle": [],
"last": "Diederik",
"suffix": ""
},
{
"first": "Jimmy",
"middle": [],
"last": "Kingma",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Ba",
"suffix": ""
}
],
"year": 2014,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1412.6980"
]
},
"num": null,
"urls": [],
"raw_text": "Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "Captum: A unified and generic model interpretability library for pytorch",
"authors": [
{
"first": "Narine",
"middle": [],
"last": "Kokhlikyan",
"suffix": ""
},
{
"first": "Vivek",
"middle": [],
"last": "Miglani",
"suffix": ""
},
{
"first": "Miguel",
"middle": [],
"last": "Martin",
"suffix": ""
},
{
"first": "Edward",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Bilal",
"middle": [],
"last": "Alsallakh",
"suffix": ""
},
{
"first": "Jonathan",
"middle": [],
"last": "Reynolds",
"suffix": ""
},
{
"first": "Alexander",
"middle": [],
"last": "Melnikov",
"suffix": ""
},
{
"first": "Natalia",
"middle": [],
"last": "Kliushkina",
"suffix": ""
},
{
"first": "Carlos",
"middle": [],
"last": "Araya",
"suffix": ""
},
{
"first": "Siqi",
"middle": [],
"last": "Yan",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2009.07896"
]
},
"num": null,
"urls": [],
"raw_text": "Narine Kokhlikyan, Vivek Miglani, Miguel Martin, Edward Wang, Bilal Alsallakh, Jonathan Reynolds, Alexander Melnikov, Natalia Kliushkina, Carlos Araya, Siqi Yan, et al. 2020. Captum: A unified and generic model interpretability library for pytorch. arXiv preprint arXiv:2009.07896.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "ROFFa Romanian Twitter dataset for offensive language",
"authors": [
{
"first": "Mihai",
"middle": [],
"last": "Manolescu",
"suffix": ""
},
{
"first": "\u00c7agr\u0131",
"middle": [],
"last": "\u00c7\u00f6ltekin",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the International Conference on Recent Advances in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "895--900",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mihai Manolescu and \u00c7agr\u0131 \u00c7\u00f6ltekin. 2021. ROFF - a Romanian Twitter dataset for offensive language. In Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021), pages 895-900, Held Online. IN- COMA Ltd.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "Multi-label categorization of accounts of sexism using a neural framework",
"authors": [
{
"first": "Pulkit",
"middle": [],
"last": "Parikh",
"suffix": ""
},
{
"first": "Harika",
"middle": [],
"last": "Abburi",
"suffix": ""
},
{
"first": "Pinkesh",
"middle": [],
"last": "Badjatiya",
"suffix": ""
},
{
"first": "Radhika",
"middle": [],
"last": "Krishnan",
"suffix": ""
},
{
"first": "Niyati",
"middle": [],
"last": "Chhaya",
"suffix": ""
},
{
"first": "Manish",
"middle": [],
"last": "Gupta",
"suffix": ""
},
{
"first": "Vasudeva",
"middle": [],
"last": "Varma",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.48550/ARXIV.1910.04602"
]
},
"num": null,
"urls": [],
"raw_text": "Pulkit Parikh, Harika Abburi, Pinkesh Badjatiya, Rad- hika Krishnan, Niyati Chhaya, Manish Gupta, and Vasudeva Varma. 2019. Multi-label categorization of accounts of sexism using a neural framework.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "Stereotyping: the politics of representation",
"authors": [
{
"first": "Michael",
"middle": [],
"last": "Pickering",
"suffix": ""
}
],
"year": 2001,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael Pickering. 2001. Stereotyping: the politics of representation. Red Globe Press.",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "Multilingual offensive language identification for lowresource languages",
"authors": [
{
"first": "Tharindu",
"middle": [],
"last": "Ranasinghe",
"suffix": ""
},
{
"first": "Marcos",
"middle": [],
"last": "Zampieri",
"suffix": ""
}
],
"year": 2021,
"venue": "Transactions on Asian and Low-Resource Language Information Processing",
"volume": "21",
"issue": "",
"pages": "1--13",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tharindu Ranasinghe and Marcos Zampieri. 2021. Mul- tilingual offensive language identification for low- resource languages. Transactions on Asian and Low- Resource Language Information Processing, 21(1):1- 13.",
"links": null
},
"BIBREF28": {
"ref_id": "b28",
"title": "Sentence-BERT: Sentence embeddings using Siamese BERTnetworks",
"authors": [
{
"first": "Nils",
"middle": [],
"last": "Reimers",
"suffix": ""
},
{
"first": "Iryna",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)",
"volume": "",
"issue": "",
"pages": "3982--3992",
"other_ids": {
"DOI": [
"10.18653/v1/D19-1410"
]
},
"num": null,
"urls": [],
"raw_text": "Nils Reimers and Iryna Gurevych. 2019. Sentence- BERT: Sentence embeddings using Siamese BERT- networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natu- ral Language Processing (EMNLP-IJCNLP), pages 3982-3992, Hong Kong, China. Association for Com- putational Linguistics.",
"links": null
},
"BIBREF29": {
"ref_id": "b29",
"title": "Overview of exist 2021: sexism identification in social networks",
"authors": [
{
"first": "Francisco",
"middle": [],
"last": "Rodr\u00edguez-S\u00e1nchez",
"suffix": ""
},
{
"first": "Jorge",
"middle": [],
"last": "Carrillo-De Albornoz",
"suffix": ""
},
{
"first": "Laura",
"middle": [],
"last": "Plaza",
"suffix": ""
},
{
"first": "Julio",
"middle": [],
"last": "Gonzalo",
"suffix": ""
},
{
"first": "Paolo",
"middle": [],
"last": "Rosso",
"suffix": ""
},
{
"first": "Miriam",
"middle": [],
"last": "Comet",
"suffix": ""
},
{
"first": "Trinidad",
"middle": [],
"last": "Donoso",
"suffix": ""
}
],
"year": 2021,
"venue": "Procesamiento del Lenguaje Natural",
"volume": "67",
"issue": "",
"pages": "195--207",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Francisco Rodr\u00edguez-S\u00e1nchez, Jorge Carrillo-de Al- bornoz, Laura Plaza, Julio Gonzalo, Paolo Rosso, Miriam Comet, and Trinidad Donoso. 2021. Overview of exist 2021: sexism identification in so- cial networks. Procesamiento del Lenguaje Natural, 67:195-207.",
"links": null
},
"BIBREF30": {
"ref_id": "b30",
"title": "Automatic classification of sexism in social networks: An empirical study on twitter data",
"authors": [
{
"first": "Francisco",
"middle": [],
"last": "Rodr\u00edguez-S\u00e1nchez",
"suffix": ""
},
{
"first": "Jorge",
"middle": [],
"last": "Carrillo-De Albornoz",
"suffix": ""
},
{
"first": "Laura",
"middle": [],
"last": "Plaza",
"suffix": ""
}
],
"year": 2020,
"venue": "IEEE Access",
"volume": "8",
"issue": "",
"pages": "219563--219576",
"other_ids": {
"DOI": [
"10.1109/ACCESS.2020.3042604"
]
},
"num": null,
"urls": [],
"raw_text": "Francisco Rodr\u00edguez-S\u00e1nchez, Jorge Carrillo-de Al- bornoz, and Laura Plaza. 2020. Automatic classifica- tion of sexism in social networks: An empirical study on twitter data. IEEE Access, 8:219563-219576.",
"links": null
},
"BIBREF31": {
"ref_id": "b31",
"title": "Othering political women: Online misogyny, racism and ableism towards women in public life",
"authors": [
{
"first": "Rosalynd",
"middle": [],
"last": "Southern",
"suffix": ""
},
{
"first": "Emily",
"middle": [],
"last": "Harmer",
"suffix": ""
}
],
"year": 2019,
"venue": "Online Othering",
"volume": "",
"issue": "",
"pages": "187--210",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rosalynd Southern and Emily Harmer. 2019. Other- ing political women: Online misogyny, racism and ableism towards women in public life. In Online Othering, pages 187-210. Springer.",
"links": null
},
"BIBREF32": {
"ref_id": "b32",
"title": "Asynchronous pipeline for processing huge corpora on medium to low resource infrastructures",
"authors": [
{
"first": "Pedro Javier Ortiz",
"middle": [],
"last": "Su\u00e1rez",
"suffix": ""
},
{
"first": "Beno\u00eet",
"middle": [],
"last": "Sagot",
"suffix": ""
},
{
"first": "Laurent",
"middle": [],
"last": "Romary",
"suffix": ""
}
],
"year": 2019,
"venue": "7th Workshop on the Challenges in the Management of Large Corpora (CMLC-7). Leibniz-Institut f\u00fcr Deutsche Sprache",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Pedro Javier Ortiz Su\u00e1rez, Beno\u00eet Sagot, and Laurent Romary. 2019. Asynchronous pipeline for process- ing huge corpora on medium to low resource infras- tructures. In 7th Workshop on the Challenges in the Management of Large Corpora (CMLC-7). Leibniz- Institut f\u00fcr Deutsche Sprache.",
"links": null
},
"BIBREF33": {
"ref_id": "b33",
"title": "The online disinhibition effect",
"authors": [
{
"first": "John",
"middle": [],
"last": "Suler",
"suffix": ""
}
],
"year": 2004,
"venue": "Cyberpsychology & behavior : the impact of the Internet, multimedia and virtual reality on behavior and society",
"volume": "7",
"issue": "",
"pages": "321--327",
"other_ids": {
"DOI": [
"10.1089/1094931041291295"
]
},
"num": null,
"urls": [],
"raw_text": "John Suler. 2004. The online disinhibition effect. Cy- berpsychology & behavior : the impact of the Inter- net, multimedia and virtual reality on behavior and society, 7:321-6.",
"links": null
},
"BIBREF34": {
"ref_id": "b34",
"title": "Axiomatic attribution for deep networks",
"authors": [
{
"first": "Mukund",
"middle": [],
"last": "Sundararajan",
"suffix": ""
},
{
"first": "Ankur",
"middle": [],
"last": "Taly",
"suffix": ""
},
{
"first": "Qiqi",
"middle": [],
"last": "Yan",
"suffix": ""
}
],
"year": 2017,
"venue": "International conference on machine learning",
"volume": "",
"issue": "",
"pages": "3319--3328",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mukund Sundararajan, Ankur Taly, and Qiqi Yan. 2017. Axiomatic attribution for deep networks. In Interna- tional conference on machine learning, pages 3319- 3328. PMLR.",
"links": null
},
"BIBREF35": {
"ref_id": "b35",
"title": "Parallel data, tools and interfaces in opus",
"authors": [
{
"first": "J\u00f6rg",
"middle": [],
"last": "Tiedemann",
"suffix": ""
}
],
"year": 2012,
"venue": "Lrec",
"volume": "2012",
"issue": "",
"pages": "2214--2218",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J\u00f6rg Tiedemann. 2012. Parallel data, tools and inter- faces in opus. In Lrec, volume 2012, pages 2214- 2218. Citeseer.",
"links": null
},
"BIBREF36": {
"ref_id": "b36",
"title": "Predictors of anonymous cyber aggression: The role of adolescents' beliefs about anonymity, aggression, and the permanency of digital content",
"authors": [
{
"first": "Michelle",
"middle": [],
"last": "Wright",
"suffix": ""
}
],
"year": 2014,
"venue": "Cyberpsychology, behavior and social networking",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.1089/cyber.2013.0457"
]
},
"num": null,
"urls": [],
"raw_text": "Michelle Wright. 2014. Predictors of anonymous cyber aggression: The role of adolescents' beliefs about anonymity, aggression, and the permanency of dig- ital content. Cyberpsychology, behavior and social networking, 17.",
"links": null
},
"BIBREF37": {
"ref_id": "b37",
"title": "The associations between cyberbullying and callous-unemotional traits among adolescents: The moderating effect of online disinhibition",
"authors": [
{
"first": "Michelle",
"middle": [],
"last": "Wright",
"suffix": ""
},
{
"first": "Bridgette",
"middle": [],
"last": "Harper",
"suffix": ""
},
{
"first": "Sebastian",
"middle": [],
"last": "Wachs",
"suffix": ""
}
],
"year": 2018,
"venue": "Personality and Individual Differences",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.1016/j.paid.2018.04.001"
]
},
"num": null,
"urls": [],
"raw_text": "Michelle Wright, Bridgette Harper, and Sebastian Wachs. 2018. The associations between cyberbul- lying and callous-unemotional traits among adoles- cents: The moderating effect of online disinhibition. Personality and Individual Differences, 140.",
"links": null
},
"BIBREF38": {
"ref_id": "b38",
"title": "Arabic sexist comments detection in youtube: A context-aware opinion analysis approach",
"authors": [
{
"first": "Jihad",
"middle": [],
"last": "Zahir",
"suffix": ""
},
{
"first": "Youssef",
"middle": [],
"last": "Mehdi Oukaja",
"suffix": ""
},
{
"first": "Oumayma",
"middle": [
"El"
],
"last": "Ansari",
"suffix": ""
}
],
"year": 2020,
"venue": "International Congress on Information and Communication Technology",
"volume": "",
"issue": "",
"pages": "461--469",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jihad Zahir, Youssef Mehdi Oukaja, and Oumayma El Ansari. 2020. Arabic sexist comments detection in youtube: A context-aware opinion analysis approach. In International Congress on Information and Com- munication Technology, pages 461-469. Springer.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"text": "Keyness scores for the words from ROMFEM-POL corpus.",
"type_str": "figure",
"num": null,
"uris": null
},
"TABREF1": {
"html": null,
"num": null,
"text": "",
"type_str": "table",
"content": "<table><tr><td colspan=\"3\">: Samples from the ROFEMPOL corpus contain-</td></tr><tr><td>ing sexual insults</td><td/><td/></tr><tr><td>NU A FOST IUBITA? Daca</td><td colspan=\"2\">prostituata normal!</td></tr><tr><td>a fost</td><td/><td/></tr><tr><td>EN: WASN'T SHE LOVED?</td><td colspan=\"2\">prostitute it's normal!</td></tr><tr><td>If she was a</td><td/><td/></tr><tr><td>Are dreptate</td><td>fufa</td><td>asta.</td></tr><tr><td>EN: This</td><td colspan=\"2\">philanderer is right.</td></tr><tr><td>iti rup gatul instant. . .</td><td>curva</td><td>tradatoare de neam si tara</td></tr><tr><td>EN: I'll break your neck in-</td><td>whore</td><td>nation and country traitor</td></tr><tr><td>stantly. . . you</td><td/><td/></tr><tr><td>E iubita. . . A fost Basescu's</td><td>BITCH</td><td>O ordinara imputita. . .</td></tr><tr><td>EN: She is the lover. . . She</td><td>BITCH</td><td>A stinking lowlife. . .</td></tr><tr><td>was Basescu's</td><td/><td/></tr><tr><td>Vad ca pe o</td><td>tarfa</td><td>pesedista o cam deranjeaza</td></tr><tr><td>EN: I can see that a</td><td>whore</td><td>of PSD is a little bothered</td></tr><tr><td>Da,</td><td colspan=\"2\">Matracuca lui Pandele</td></tr><tr><td>EN: Yes,</td><td colspan=\"2\">the bimbo of Pandele</td></tr><tr><td>Iar tu vei face iar pus, c\u0203rie,</td><td>zdreant , o</td><td/></tr><tr><td>EN: You will go to jail,</td><td>slut</td><td/></tr></table>"
},
"TABREF2": {
"html": null,
"num": null,
"text": "Samples from the ROFEMPOL corpus containing stereotypes portrayed as \"Women's jobs\"",
"type_str": "table",
"content": "<table><tr><td>Sa stea</td><td>acasa</td><td>sa. Si creasca copiii</td></tr><tr><td>EN: She should stay</td><td>at home</td><td>to. And to raise the children</td></tr><tr><td colspan=\"2\">minte nici c\u00e2t o bibilic\u0203!!! Stai acas\u0203</td><td>s, i cros, eteaz\u0203 nu te mai face de</td></tr><tr><td/><td/><td>rahat</td></tr><tr><td>EN: birdbrain!!! Stay</td><td>home</td><td>and knit, don't embarrass your-</td></tr><tr><td/><td/><td>self</td></tr><tr><td>acum ca e</td><td>casnica</td><td>la rosiorii de vede</td></tr><tr><td>EN: now that she is a</td><td>housewife</td><td>at rosiorii de vede</td></tr><tr><td>Eroin\u0203, dar nu s, i</td><td>mam\u0203</td><td>. . . :-(</td></tr><tr><td>EN: A hero, but not a</td><td>mother</td><td>. . . :-(</td></tr><tr><td>Acum totul e o nebunie. . . tu</td><td>mam\u0203</td><td>sot , ie lasa i naibii de politicieni</td></tr><tr><td>esti</td><td/><td>si vezi</td></tr><tr><td>EN: Now everything is insane. . .</td><td>mother</td><td>a wife, forget those goddamn</td></tr><tr><td>you are a</td><td/><td>politicians</td></tr><tr><td>dar sunt sigur ca si in</td><td>bucatarie</td><td>le mai incurci.</td></tr><tr><td>EN: but I am sure that in</td><td>the kitchen</td><td>you also mess up sometimes.</td></tr><tr><td>Apuc\u0203-te Vasilico de</td><td>sarmalele</td><td>alea, las\u0203 basmele!</td></tr><tr><td>EN: Vasilico, start cooking</td><td colspan=\"2\">cabbage rolls , forget the fairytales!</td></tr><tr><td>those</td><td/><td/></tr><tr><td>O fi terminat de f\u0203cut</td><td>zacusca</td><td>s, i acum se plictises, te?!</td></tr><tr><td>EN: Maybe she finished making</td><td>zacusca</td><td>and now she is bored?!</td></tr><tr><td>the</td><td/><td/></tr><tr><td>doamne ce moaca de</td><td>bucatareasa</td><td>nepregatita are, un jeg, un gunoi</td></tr><tr><td>EN: god, she looks like a un-</td><td>cook</td><td>, dirtbag, she is garbage</td></tr><tr><td>skilled</td><td/><td/></tr><tr><td/><td>Spalatoreasa</td><td>asta il ironineaza pe Klaus</td></tr><tr><td>EN: This</td><td colspan=\"2\">washerwomen mocks Klaus</td></tr><tr><td>grapini zici ca e</td><td>femeia de ser-</td><td>acolo, are grija sa nu se faca</td></tr><tr><td/><td>vici</td><td>mizerie</td></tr><tr><td>EN: grapini looks like</td><td>the cleaning</td><td>there, she takes care not to make</td></tr><tr><td/><td>lady</td><td>a mess</td></tr><tr><td>o rapandula</td><td>secretara</td><td>din videle este adevarat-ce</td></tr><tr><td/><td/><td>spune</td></tr><tr><td>EN: A slut,</td><td/><td/></tr></table>"
},
"TABREF3": {
"html": null,
"num": null,
"text": "Samples from the ROFEMPOL corpus containing stereotypes portrayed as \"Female attitudes\"",
"type_str": "table",
"content": "<table><tr><td>Eu tot o spun, o</td><td>femeie</td><td>nu are un rol activ</td></tr><tr><td>EN: I keep saying it, a</td><td>woman</td><td>doesn't have an active role</td></tr><tr><td colspan=\"2\">sa fie condus\u0203 de femeie, numai o femeie</td><td>poate avea grija de cet\u0203t , enii ei pre-</td></tr><tr><td/><td/><td>cum</td></tr><tr><td colspan=\"2\">EN: to be led by a woman, only a woman</td><td>can take care of the citizens like</td></tr><tr><td colspan=\"2\">EA insasi fiinta, parte din feminin, matern</td><td>umanitate, parte din istoria</td></tr><tr><td>EN: SHE is being itself, part of</td><td>maternal</td><td>humanity, part of history</td></tr><tr><td>the feminine,</td><td/><td/></tr><tr><td>USRist</td><td>isteric\u0203</td><td>prins\u0203 cu m\u00e2t , a-n sacul de voturi?</td></tr><tr><td>EN: The USRite is</td><td colspan=\"2\">hysterical when caught with her hand in the</td></tr><tr><td/><td/><td>vote jar?</td></tr><tr><td>Femeia este precum o piranda</td><td colspan=\"2\">nervoasa dispusa oricand sa-si ridice</td></tr><tr><td>EN: The woman is like a gypsy</td><td>angry</td><td>ready at any time to draw up</td></tr><tr><td>who is</td><td/><td/></tr><tr><td>trebuie sa munceasc\u0203 nu sa</td><td colspan=\"2\">b\u00e2rfeasc\u0203 Firea a fost numit\u0203</td></tr><tr><td>EN: she must work, not</td><td>gossip</td><td>Firea was named</td></tr><tr><td>D-na Grapini era singura s, i cea</td><td>crizat\u0203</td><td>doamn\u0203 din sal\u0203. M\u0203 bucur c\u0203 a</td></tr><tr><td>mai</td><td/><td/></tr><tr><td>EN: D-na Grapini was the only</td><td colspan=\"2\">hysterical lady in the room.</td></tr><tr><td>and most</td><td/><td/></tr></table>"
},
"TABREF4": {
"html": null,
"num": null,
"text": "Samples from the ROFEMPOL corpus containing mocking female titles",
"type_str": "table",
"content": "<table><tr><td/><td>Cucoana</td><td>cplm te recomanda sa fii primar</td></tr><tr><td>EN:</td><td>Lady</td><td>what the hell recommends you</td></tr><tr><td/><td/><td>to be mayor</td></tr><tr><td/><td>Tanti</td><td>Nuti,daca ai spune adevarul, tu</td></tr><tr><td/><td/><td>si</td></tr><tr><td>EN:</td><td>Auntie</td><td>Nuti, if you told the truth</td></tr><tr><td colspan=\"2\">de unde are at\u00e2ta tupeu aceast\u0203 madam\u0203</td><td/></tr><tr><td>EN: where does this</td><td>madame</td><td>get the nerve</td></tr><tr><td colspan=\"2\">ce Marete Realizari a inregistrat DUamna</td><td>Gabi? Cum a inceput Macar</td></tr><tr><td>EN: what Great Achievements</td><td>LAdy</td><td>Gabi have? How did she start</td></tr><tr><td>doe</td><td/><td/></tr><tr><td>Cui ii mai pasa de</td><td>domnisoara</td><td>batrana? Sa stea acolo fara</td></tr><tr><td>EN: Who cares about this</td><td>spinster</td><td>? She should stay there without</td></tr><tr><td>Bravo</td><td>fata</td><td>sper sa fii ca o pum\u0203</td></tr><tr><td>EN: Well done</td><td>girl</td><td>I hope you will be like a puma</td></tr><tr><td>nu inteleg cine a pus-o pe</td><td>fetiscana</td><td>sa candideze ca primar</td></tr><tr><td>aceasta</td><td/><td/></tr><tr><td>EN: I don't understand who</td><td>schoolgirl</td><td>run for mayor</td></tr><tr><td>made this</td><td/><td/></tr><tr><td>Viata va fi foarte grea pentru</td><td colspan=\"2\">ex-ministruoasa si cei 3-4 iubiti</td></tr><tr><td>EN: Life is very hard for the</td><td>ex-ministress</td><td>and the 3-4 lovers</td></tr><tr><td>Ce aveti cu biata</td><td>ministreuz\u0203</td><td>Ea a r\u0203spuns correct.</td></tr><tr><td>EN: What do you have against</td><td>ministreaseuse</td><td>She answered corrrectly.</td></tr><tr><td>the poor</td><td/><td/></tr><tr><td>Doamna</td><td>prim\u0203rit ,\u0203</td><td>tace s, i face! Si-a inceput man-</td></tr><tr><td/><td/><td>datul</td></tr><tr><td>EN: Madam</td><td>mayoress</td><td>gets the job done!</td></tr><tr><td>d) Female Titles</td><td/><td/></tr></table>"
},
"TABREF5": {
"html": null,
"num": null,
"text": "Samples from the ROFEMPOL corpus containing body shaming",
"type_str": "table",
"content": "<table><tr><td>Ca vrea cineva sao imbol-</td><td>grasa</td><td>asta? E cantitate nesemni-</td></tr><tr><td>naveasca pe</td><td/><td>ficativa.</td></tr><tr><td colspan=\"2\">EN: Who wants to get this fatso</td><td>sick? She doesn't matter.</td></tr><tr><td>un</td><td>porc</td><td>de o asemenea greutate</td></tr><tr><td>EN: a</td><td>pig</td><td>of this weight</td></tr><tr><td>in</td><td colspan=\"2\">matahala asta de Sosoaca isi gasesc</td></tr><tr><td/><td/><td>modelul</td></tr><tr><td>EN: they see this</td><td>hulk</td><td>Sosoaca as a model</td></tr></table>"
},
"TABREF6": {
"html": null,
"num": null,
"text": "Samples from the ROFEMPOL corpus containing hair color stereotypes",
"type_str": "table",
"content": "<table><tr><td/><td colspan=\"2\">Blondo vezi ca ies , im in strada</td></tr><tr><td>EN:</td><td colspan=\"2\">Blondie we will riot</td></tr><tr><td>N\u0203p\u00e2rc\u0203</td><td/><td>ce vrea s\u0203 par\u0203 in-</td></tr><tr><td/><td>blond\u0203</td><td>ocent\u0203(!)</td></tr><tr><td>EN: A</td><td colspan=\"2\">blonde adder who wants to look</td></tr><tr><td/><td/><td>innocent(!)</td></tr></table>"
},
"TABREF7": {
"html": null,
"num": null,
"text": "Samples from the ROFEMPOL corpus containing sexists comments related to clothes and style",
"type_str": "table",
"content": "<table><tr><td>Mana Iute a imbracat in</td><td colspan=\"2\">rochie murdara tesuta cu mo-</td></tr><tr><td>campanie o</td><td/><td>tive romanesti..</td></tr><tr><td>EN: Sticky Fingers</td><td colspan=\"2\">dress that was dirty and with</td></tr><tr><td>wore, during the cam-</td><td/><td>traditional motifs</td></tr><tr><td>paign, a</td><td/><td/></tr><tr><td>p\u0203i puteai s\u0203 stai</td><td>fust\u0203</td><td>scurt\u0203 s , i cu fundul pe</td></tr><tr><td/><td/><td>birou as , a</td></tr><tr><td>EN: you could sit in a</td><td>skirt</td><td>and with your ass on the</td></tr><tr><td>short</td><td/><td>desk</td></tr><tr><td>valuta neagra spalata in</td><td colspan=\"2\">haine si accesorii de sute de</td></tr><tr><td>tara,</td><td/><td>mii de euro</td></tr><tr><td>EN: foreign currency</td><td colspan=\"2\">clothes and accessories worth</td></tr><tr><td>laundered in the coun-</td><td/><td>thousands of euros</td></tr><tr><td>try,</td><td/><td/></tr></table>"
},
"TABREF8": {
"html": null,
"num": null,
"text": "",
"type_str": "table",
"content": "<table><tr><td colspan=\"4\">: Samples from the ROFEMPOL corpus con-</td></tr><tr><td colspan=\"4\">taining sexists comments related to perceived physical</td></tr><tr><td>attractiveness</td><td/><td/></tr><tr><td>mai vedem aceasta agra-</td><td>urata</td><td>femeie</td></tr><tr><td>mata</td><td/><td/></tr><tr><td>EN: can we still see this</td><td>ugly</td><td>woman</td></tr><tr><td>illiterate woman? an</td><td/><td/></tr><tr><td/><td colspan=\"2\">Hidos , enia trebuie</td><td>mascat\u0203</td></tr><tr><td/><td/><td>cumva. . .</td></tr><tr><td>EN:</td><td colspan=\"3\">Hideousness must be hidden some-</td></tr><tr><td/><td/><td>how</td></tr><tr><td>dumneata esti, si ai fost</td><td>frumoasa</td><td colspan=\"2\">:) si chiar: inca foarte</td></tr><tr><td>o papusa</td><td/><td>frumoasa :)!</td></tr><tr><td>EN: you are, and have</td><td>beautiful</td><td colspan=\"2\">doll :) and even: still</td></tr><tr><td>always been a</td><td/><td colspan=\"2\">very beautiful :)!</td></tr><tr><td>NU E</td><td>FEMEIE</td><td colspan=\"2\">De-aia au si ales-o devi-</td></tr><tr><td/><td/><td>atii de la</td></tr><tr><td>EN: she IS NOT A</td><td>WOMAN</td><td colspan=\"2\">That's why the deviants</td></tr><tr><td/><td/><td colspan=\"2\">have chosen her</td></tr></table>"
},
"TABREF9": {
"html": null,
"num": null,
"text": "",
"type_str": "table",
"content": "<table><tr><td colspan=\"3\">: Samples from the ROFEMPOL corpus contain-</td></tr><tr><td colspan=\"2\">ing dehumanisation language</td><td/></tr><tr><td>Baaa, v-ati uitat bine la</td><td colspan=\"2\">creatura asta? Dupa cum se imbraca</td></tr><tr><td>EN: Maaan, have you looked</td><td colspan=\"2\">creature ? Judging by the way she</td></tr><tr><td>carefully at this</td><td/><td>dresses</td></tr><tr><td>Bai</td><td>vaca</td><td>descreierata, tu neaparat ai</td></tr><tr><td/><td/><td>nevoie</td></tr><tr><td>EN: You brainless</td><td>cow</td><td>, you definitely need</td></tr><tr><td>La puscarie cu tine,</td><td>javra</td><td>muista ordinara!</td></tr><tr><td>EN: Rot in jail,</td><td>bitch</td><td>disgusting cocksucker!</td></tr><tr><td>cu g\u00e2ndul la... os. Este ca o</td><td>c\u0203t , ea</td><td>\u00een c\u0203lduri</td></tr><tr><td>EN: always thinks of... bon-</td><td>bitch</td><td>in heat</td></tr><tr><td>ing. She is like a</td><td/><td/></tr><tr><td>Sau t , ie \u00eet , i place gunoiul,</td><td colspan=\"2\">scroaf\u0203 alogen\u0203!</td></tr><tr><td>EN: Or you like garbage, you</td><td>sow</td><td>!</td></tr><tr><td>foreign</td><td/><td/></tr></table>"
},
"TABREF10": {
"html": null,
"num": null,
"text": "68\u00b10.01 0.78\u00b10.02 0.72\u00b10.01 0.66\u00b10.02 0.52\u00b10.02 0.58\u00b10.01 0.67\u00b10.01 0.67\u00b10.01 0.66\u00b10.01 0.66\u00b10.01",
"type_str": "table",
"content": "<table><tr><td/><td/><td>Non-sexist</td><td/><td/><td>Sexist</td><td/><td colspan=\"2\">Weighted Average</td><td/><td/></tr><tr><td>Model</td><td>Precision</td><td>Recall</td><td>F1</td><td>Precision</td><td>Recall</td><td>F 1</td><td>Precision</td><td>Recall</td><td>F 1</td><td>Macro-F 1</td></tr><tr><td colspan=\"11\">BOW + LR 0.BOW + RF 0.62\u00b10.01 0.90\u00b10.05 0.74\u00b10.03 0.72\u00b10.10 0.31\u00b10.02 0.43\u00b10.03 0.67\u00b10.05 0.64\u00b10.03 0.60\u00b10.02 0.58\u00b10.02</td></tr><tr><td>BOW + SVM</td><td colspan=\"10\">0.68\u00b10.02 0.79\u00b10.01 0.72\u00b10.01 0.65\u00b10.02 0.50\u00b10.04 0.56\u00b10.03 0.66\u00b10.01 0.66\u00b10.02 0.65\u00b10.02 0.64\u00b10.02</td></tr><tr><td>TFIDF + LR</td><td colspan=\"10\">0.70\u00b10.01 0.79\u00b10.01 0.74\u00b10.00 0.68\u00b10.00 0.56\u00b10.03 0.61\u00b10.02 0.69\u00b10.00 0.69\u00b10.01 0.68\u00b10.01 0.68\u00b10.01</td></tr><tr><td>TFIDF + RF</td><td colspan=\"10\">0.63\u00b10.01 0.90\u00b10.04 0.74\u00b10.02 0.72\u00b10.06 0.31\u00b10.02 0.44\u00b10.02 0.66\u00b10.03 0.64\u00b10.02 0.61\u00b10.01 0.59\u00b10.02</td></tr><tr><td>TFIDF + SVM</td><td colspan=\"10\">0.69\u00b10.01 0.77\u00b10.02 0.73\u00b10.01 0.66\u00b10.01 0.56\u00b10.03 0.61\u00b10.02 0.68\u00b10.01 0.68\u00b10.01 0.68\u00b10.01 0.67\u00b10.01</td></tr><tr><td>M-BERT emb + LR</td><td colspan=\"10\">0.68\u00b10.01 0.80\u00b10.01 0.74\u00b10.00 0.67\u00b10.01 0.53\u00b10.01 0.59\u00b10.01 0.68\u00b10.01 0.68\u00b10.01 0.67\u00b10.01 0.66\u00b10.01</td></tr><tr><td>M-BERT emb + RF</td><td colspan=\"10\">0.64\u00b10.00 0.82\u00b10.02 0.72\u00b10.01 0.64\u00b10.02 0.41\u00b10.01 0.50\u00b10.01 0.64\u00b10.01 0.64\u00b10.01 0.62\u00b10.00 0.61\u00b10.01</td></tr><tr><td colspan=\"11\">M-BERT emb + SVM 0.67\u00b10.01 0.79\u00b10.02 0.72\u00b10.01 0.65\u00b10.02 0.50\u00b10.04 0.57\u00b10.02 0.66\u00b10.01 0.66\u00b10.01 0.66\u00b10.01 0.64\u00b10.01</td></tr><tr><td>Ro-BERT emb + LR</td><td colspan=\"10\">0.70\u00b10.01 0.78\u00b10.00 0.74\u00b10.01 0.67\u00b10.01 0.57\u00b10.03 0.62\u00b10.01 0.69\u00b10.01 0.69\u00b10.01 0.69\u00b10.01 0.68\u00b10.01</td></tr><tr><td>Ro-BERT emb + RF</td><td colspan=\"10\">0.68\u00b10.01 0.83\u00b10.01 0.74\u00b10.01 0.69\u00b10.02 0.49\u00b10.03 0.57\u00b10.02 0.68\u00b10.01 0.68\u00b10.02 0.67\u00b10.02 0.66\u00b10.02</td></tr><tr><td colspan=\"11\">Ro-BERT emb + SVM 0.71\u00b10.01 0.79\u00b10.02 0.75\u00b10.01 0.69\u00b10.02 0.58\u00b10.02 0.63\u00b10.01 0.70\u00b10.01 0.70\u00b10.01 0.70\u00b10.01 0.69\u00b10.01</td></tr><tr><td>Fine-tuned M-BERT</td><td colspan=\"10\">0.77\u00b10.05 0.75\u00b10.02 0.76\u00b10.02 0.67\u00b10.06 0.70\u00b10.03 0.69\u00b10.02 0.74\u00b10.02 0.73\u00b10.01 0.73\u00b10.01 0.72\u00b10.01</td></tr><tr><td>Fine-tuned Ro-BERT</td><td colspan=\"10\">0.76\u00b10.05 0.80\u00b10.03 0.78\u00b10.01 0.76\u00b10.06 0.71\u00b10.03 0.73\u00b10.02 0.77\u00b10.01 0.76\u00b10.01 0.76\u00b10.01 0.75\u00b10.01</td></tr></table>"
},
"TABREF11": {
"html": null,
"num": null,
"text": "Selected correct predictions using Ro-BERT, the best performing model.",
"type_str": "table",
"content": "<table><tr><td>Non-sexist</td><td>Sexist</td></tr><tr><td>Asa ai timp de lectura. C\u00e2nd</td><td>vrea si iea la festivalul homosex-</td></tr><tr><td>erai Primar nu era timp deloc</td><td>ualilor in sibiu? jigodie pdlista</td></tr><tr><td/><td>vinzatore de tzara</td></tr><tr><td>EN: This way, you have time to</td><td>does she want to go to the ho-</td></tr><tr><td>read. When you were Mayor,</td><td>mosexual festival in sibiu? PDL</td></tr><tr><td>you didn't have time at all</td><td>bitch, country traitor</td></tr><tr><td>un primar asa tot sa avem bravo</td><td>C\u0203t , eaua asta trebuie dus\u0203 acolo</td></tr><tr><td>t , ie olguta .un primar ambitios!</td><td>\u00eei este locul -Jilava!</td></tr><tr><td>EN: we are glad to have such</td><td>This bitch must be taken where</td></tr><tr><td>a mayor well done olguta .an</td><td>she belongs -Jilava!</td></tr><tr><td>ambitious mayor</td><td/></tr><tr><td>Doamna Olgut , a trebuie sa-s , i</td><td>Veo, n-ai mur\u0203turi de pus?</td></tr><tr><td>sporeasc\u0203 averea.</td><td/></tr><tr><td>EN: Mrs. Olgut , a has to increase</td><td>Veo, don't you have pickles to</td></tr><tr><td>her fortune.</td><td>make?</td></tr></table>"
},
"TABREF12": {
"html": null,
"num": null,
"text": "Selected incorrect predictions using Ro-BERT.",
"type_str": "table",
"content": "<table><tr><td>Prediction: Non-sexist</td><td>Prediction: Sexist</td></tr><tr><td>True Label: Sexist</td><td>True Label: Non-sexist</td></tr><tr><td>Hudrea tot pe centura . . . .</td><td>A fost o data o Diana print , esa</td></tr><tr><td/><td>a inimilor acum Dumnezeu ne</td></tr><tr><td/><td>a trimes o Diana print , es\u0203 a</td></tr><tr><td/><td>Rom\u00e2nilor</td></tr><tr><td>EN: Hudrea keeps working the</td><td>There once was a Diana,</td></tr><tr><td>streets . . . .</td><td>princess of hearts now God has</td></tr><tr><td/><td>sent us a Diana princess of the</td></tr><tr><td/><td>Romanians</td></tr><tr><td>Ai reusit deja tuto, sa distrugi</td><td>Clocosoros asta,e imaginea</td></tr><tr><td>o generatie din viitorul Ro-</td><td>apocalipsei,intruchiparea</td></tr><tr><td>maniei!.</td><td>raului,arata ca un paznic la</td></tr><tr><td/><td>poarta Infernului. . . . O meritati</td></tr><tr><td/><td>cretinilor !</td></tr><tr><td>EN: You have already managed,</td><td>This Clocosoros is the image</td></tr><tr><td>dumbass, to destroy a genera-</td><td>of the apocalypse, the embod-</td></tr><tr><td>tion of the future of Romania!.</td><td>iment of evil, she looks like a</td></tr><tr><td/><td>guard at the gate of hell. . . You</td></tr><tr><td/><td>deserve her, idiots!</td></tr><tr><td>Felicitari pentru articol . . . .o</td><td>Analfabet\u0203, incult\u0203 , scursura</td></tr><tr><td>TOAPA parvenita si needucata</td><td>societ\u0203t , ii ...</td></tr><tr><td>!</td><td/></tr><tr><td>EN: Congratulations on the ar-</td><td>You illiterate uneducated trash</td></tr><tr><td>ticle. . . An uneducated boorish</td><td>of society. . .</td></tr><tr><td>upstart!</td><td/></tr></table>"
}
}
}
}