Volume 19, Issue 2 - Languages of offence in digital perspectives, Guest-edited by Barbara Lewandowska-Tomaszczyk and Marcin Trojszczak

This study intends to contribute to the delimitation of selected offensive language categories based on an analysis of a corpus of contributions to discussion forums in Czech online national newspapers and news platforms called Czech Corpus of Offensive Language (CCOL). It endeavours to study three problematic areas (1) delimitation between the speech acts performed, (ii) lexical realisation of specific properties of the target and (iii) identification and categorisation of implicit offence (e.g. figurative semantic shifts) by exploring contextual cues for the speech act identification, the keywords indicating the properties of the target and the types of semantic shifts in implicit expressions of offence. The findings indicate that annotation systems that do not use context information for the detection of offensive language may face problems with adequate interpretation of the language means under investigation.

The aim of the current research is to investigate the feasibility of identifying offensive language in Lithuanian by utilising the Simplified Offensive Language Taxonomy (SOLT). The key principle behind this taxonomy is its ability to complement existing offensive language ontologies and tagset systems, with the ultimate goal of integrating it into publicly accessible Linguistic Linked Open Data (LLOD) resources. The dataset used in the current study is a publicly available corpus of user-generated comments collected from a Lithuanian portal (Amilevičius et al. 2016). The study identified that offensive language predominantly focuses on collective derogatory language rather than individuals. The most common category of offensive language is related to physical and mental disabilities, followed by ideological offenses, xenophobic and sexist remarks, and less frequent categories like ageism, classism, homophobia, and religious discrimination. These results highlight the diverse range of offensive language online and underscore the need to combat discrimination and promote respectful discourse, particularly concerning marginalised groups.

This paper intends to examine the development of conflictual interactions, how they might be resolved, and the socio-cultural norms involved, by adopting an analytical framework in an online gaming context. The current paper was inspired by Kádár and Haugh’s framework as it enables me to investigate both the macro and micro aspects of (im)politeness. The study’s aim is to further examine how impoliteness, language aggression and conflict are realised in two online gaming platforms, namely Fortnite and PUBG Mobile. Thus, I will explore discursively how these phenomena are subjective in a Tunisian Arabic setting and discuss how participants reach their subjective perception of conflict in ways that do not always correspond to the supposed intentions of the ostensible offender. The results indicate that conflict is subjective as it is evaluated in different ways by different gamers and could be a result of a conflictual intention. The findings also reveal that conflict may also be created/ escalated as a result of a non-conflictual intention. Thus, this paper contributes to understanding of conflict, how impoliteness can lead to conflict and the various aspects of impoliteness/the perception of impoliteness.

This contribution takes a look at video-sharing platforms to highlight a popular entertainment format which consists in re-framing political discourse for the purposes of entertaining the audience and, at the same time, providing an evaluation of that discourse. Evaluations of political discourse uncover the role and importance imputed to it by those who are outside of the political system, but who are directly impacted by it, that is, the people. A sample of French-language data, collected from YouTube, is examined for the categories of evaluation which are used by the authors, which are conjectured to represent the ideas ordinary citizens have about political discourse. Reframing political discourse carries these evaluations through offensive language and humor. Teasing out the relevant parameters of evaluation can provide a basis for understanding how regular citizens gauge political discourse. Appropriateness appears to be an important evaluation criterion pertaining to the characteristics of political discourse, occurring in a particular context.

The goal of the paper is to present a Simplified Offensive Language (SOL) Taxonomy, its application and testing in the Second Annotation Campaign conducted between March-May 2023 on four languages: English, Czech, Lithuanian, and Polish to be verified and located in LLOD. Making reference to the previous Offensive Language taxonomic models proposed mostly by the same COST Action Nexus Linguarum WG 4.1.1 team, the number and variety of the categories underwent the definitional revision, and the present typology was tested in the annotation on the publicly available offensive language datasets of each of the four languages. The results of the annotation are presented and as they are contained within the accepted statistical values on the inter-annotator agreement in the SOL categories and their aspects, we propose this taxonomy as a core ontology which represents the encoding of the supported offensive languages and justify its use on new data in terms of a more universal Linguistic Linked Open Data (LLOD) schema.

This paper introduces a streamlined taxonomy for categorizing offensive language in Hebrew, addressing a gap in the literature that has, until now, largely focused on Indo-European languages. Our taxonomy divides offensive language into seven levels (six explicit and one implicit level). We based our work on the simplified offensive language (SOL) taxonomy introduced in (Lewandowska-Tomaszczyk et al. 2021a) hoping that our adjustment of SOL to the Hebrew language will be capable of reflecting the unique linguistic and cultural nuances of Hebrew. The study involves both linguistic and cultural analysis beyond Natural Language Processing (NLP). We employed manual linguistic analysis to understand the nuances of offensive language in Hebrew. An accompanying dataset, gathered on Twitter and manually curated by human annotators, is described in detail. This dataset was constructed to both validate the taxonomy and serve as a foundation for future research on offensive language detection and analysis in Hebrew. Preliminary analysis of the dataset reveals intriguing patterns and distributions, underscoring the complexity and specificity of offensive expressions in the Hebrew language. The aim of our work is to capture the complexity and specificity of offensive expressions in Hebrew beyond what automated NLP methods alone can provide. Our findings highlight the significance of considering linguistic and cultural variations when researching and correcting abusive language online. We believe that our streamlined taxonomy and associated dataset will be crucial in improving research in Hebrew language sociocultural studies, natural language processing, and offensive language detection. Our study also makes a substantial contribution to the study of low-resource languages and can be used as a model for future research on other languages.

The aim of this paper is to shed some light on the linguistic concept of implicit offensiveness. On the one hand, implicitness will be juxtaposed with indirectness as the two concepts are not conceived of here as synonymous. On the other hand, a typology of offensiveness (vs offensive language and vs offendedness) will be proposed, as well as the overarching term ‘covert meaning’ that will span figurative implicitness and non-figurative implicitness. The gradability of various forms of covert meaning and its overlap with overt meaning (subsuming explicit literal meaning and implicit literal meaning) will also be discussed. In the analysis, two sample implicit concepts will be examined (irony vs sarcasm) based on corpus data (of general English and dedicated offensiveness corpus) and using non-contextual embeddings. Theory-wise, the paper demonstrates that implicitness is a complex term which is fuzzy and gradable; methodology-wise, it shows how computational tools can be used to attest theoretical assumptions related to offensive covert terms.

The automatic detection of implicitly offensive language is a challenge for NLP, as such language is subtle, contextual, and plausibly deniable, but it is becoming increasingly important with the wider use of large language models to generate human-quality texts. This study argues that current difficulties in detecting implicit offence are exacerbated by multiple factors: (a) inadequate definitions of implicit and explicit offense; (b) an insufficient typology of implicit offence; and (c) a dearth of detailed analysis of implicitly offensive linguistic data. In this study, based on a qualitative analysis of an implicitly offensive dataset, a new typology of implicitly offensive language is proposed along with a detailed, example-led account of the new typology, an operational definition of implicitly offensive language, and a thorough analysis of the role of figurative language and humour in each type. Our analyses identify three main issues with previous datasets and typologies used in NLP approaches: (a) conflating content and form in the annotation; (b) treating figurativeness, particularly metaphor, as the main device of implicitness, while ignoring its equally important role in the explicit offence; and (c) an over-focus on form-specific datasets (e.g. focusing only on offensive comparisons), which fails to reflect the full complexity of offensive language use.

Currently, the Internet information and communication network has become an integral part of human life. People use social networks such as Twitter, VKontakte, Facebook, etc., to establish global contacts, exchange opinions, gain knowledge, etc. The active participation of not only individual users, but also information organizations in the entire world space makes it necessary to develop measures that correspond to modern trends in the development of information and communication technologies to ensure national security, in particular, the organization of events related to countering the strengthening of ideas of extremism and terrorism. Countering the spread of aggressive information on the global network is an urgent problem of society and government agencies, this task is solved by filtering unwanted Internet resources. However, terrorist and extremist groups rationally use web technologies to perform various functions, including information dissemination, propaganda, fundraising and extremist missions. In such a situation, the Internet poses a threat to national security. In this paper, we investigate the issue of creating semantic analysis models to identify extremist messages in the Kazakh language. For the study, a proprietary text corpus was assembled and models based on bigrams and word input methods were proposed. According to the results of experiments, the proposed model shows the highest indicators for evaluating machine learning methods.

The prevalence of sensationalized headlines and deceptive narratives in online content has prompted the need for effective clickbait detection methods. This study delves into the nuances of clickbait in Hebrew, scrutinizing diverse features such as linguistic and structural features, and exploring various types of clickbait in Hebrew, a language that has received relatively limited attention in this context. Utilizing a range of machine learning models, this research aims to identify linguistic features that are instrumental in accurately classifying Hebrew headlines as either clickbait or non-clickbait. The findings underscore the critical role of linguistic attributes in enhancing the performance of the classification model. Notably, the employment of a machine learning model resulted in an impressive accuracy of 0.87 in clickbait detection. Moving forward, our research plan encompasses dataset expansion through the best machine learning model assisted labelling, with the objective of optimizing deep learning models for even more robust outcomes. This study not only advances clickbait detection in the realm of Hebrew but also emphasizes the fundamental importance of linguistic features in the accurate classification of clickbait.

The paper investigates various definitions of the concept of opinion as opposed to factual or evidence-based statements and proposes a taxonomy of opinions expressed in English as identified in selected social media. A discussion situates opinions in the realm of pragmatics and reaches to philosophy of language and cognitive science. The research methodology combines a thorough linguistic analysis of opinions, proposing their multifaceted taxonomy with the automatically generated lexical embeddings of positive and negative lexicon acquired from the analysed opinionated texts. As proposed, the definition of the concept of opinion is best apprehended when looked upon in terms of an opinion event , with a number of necessary conditions on the one hand, and those that are characteristic of an explicit opinion prototype on the other. Lists of opinion discourse markers show their preferential uses either in positive or negative opinionated texts; however, no sets of necessary and/or sufficient opinion markers properties have been acquired from the analysed texts. The conclusions indicate a polysemous understanding of naturally occurring social media opinionated texts and a definitional flexibility of the boundaries around lexical positive and negative types of opinion markers.