The current study represents an integrated model of explicit and implicit offensive language taxonomy. First, it focuses on a definitional revision and enrichment of the explicit offensive language taxonomy by reviewing the collection of available corpora and comparing tagging schemas applied there. The study relies mainly on the categories originally proposed by Zampieri et al. (2019) in terms of offensive language categorization schemata. After the explanation of semantic differences between particular concepts used in the tagging systems and the analysis of theoretical frameworks, a finite set of classes is presented, which cover aspects of offensive language representation along with linguistically sound explanations (Lewandowska-Tomaszczyk et al. 2021). In the analytic procedure, offensive from non-offensive discourse is first distinguished, with the question of offence Target and the following categorization levels and sublevels. Based on the relevant data generated from Sketch Engine (https://www.sketchengine.eu/ententen-english-corpus/), we propose the concept of offensive language as a superordinate category in our system with a number of hierarchically arranged 17 subcategories. The categories are taxonomically structured into 4 levels and verified with the use of neural-based (lexical) embeddings. Together with a taxonomy of implicit offensive language and its subcategorization levels which has received little scholarly attention until now, the categorization is exemplified in samples of offensive discourses in selected English social media materials, i.e., publicly available 25 web-based hate speech datasets (consult Appendix 1 for a complete list). The offensive category levels (types of offence, targets, etc.) and aspects (offensive language property clusters) as well as the categories of explicitness and implicitness are discussed in the study and the computationally verified integrated explicit and implicit offensive language taxonomy proposed in the study.

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.

In this study, we aim to classify comments as abusive or non-abusive. We develop a Hebrew corpus of user comments annotated for abusive language. Then, we investigate highly sparse n-grams representations as well as denser character n-grams representations for comment abuse classification. Since the comments in social media are usually short, we also investigate four dimension reduction methods, which produce word vectors that collapse similar words into groups. We show that the character n-grams representations outperform all the other representation for the task of identifying abusive comments.

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 article describes methods for semiautomatic thesaurus construction, for a cross generation, cross genre, and cross cultural corpus. Semiautomatic thesaurus construction is a complex task, and applying it on a cross generation corpus brings its own challenges. We used a Jewish juristic corpus containing documents and genres that were written across 2000 years, and contain a mix of different languages, dialects, geographies, and writing styles. We evaluated different first and second order methods, and introduced a special annotation scheme for this problem, which showed that first order methods performed surprisingly well. We found that in our case, improving the coverage is the more difficult task, for this we introduce a new algorithm to increase recall (coverage)—which is applicable to many other problems as well, and demonstrates significant improvement in our corpus.