Aiqi Jiang

Wenjie Yin, Vibhor Agarwal, Aiqi Jiang et al.

Supervised approaches generally rely on majority-based labels. However, it is hard to achieve high agreement among annotators in subjective tasks such as hate speech detection. Existing neural network models principally regard labels as categorical variables, while ignoring the semantic information in diverse label texts. In this paper, we propose AnnoBERT, a first-of-its-kind architecture integrating annotator characteristics and label text with a transformer-based model to detect hate speech, with unique representations based on each annotator's characteristics via Collaborative Topic Regression (CTR) and integrate label text to enrich textual representations. During training, the model associates annotators with their label choices given a piece of text; during evaluation, when label information is not available, the model predicts the aggregated label given by the participating annotators by utilising the learnt association. The proposed approach displayed an advantage in detecting hate speech, especially in the minority class and edge cases with annotator disagreement. Improvement in the overall performance is the largest when the dataset is more label-imbalanced, suggesting its practical value in identifying real-world hate speech, as the volume of hate speech in-the-wild is extremely small on social media, when compared with normal (non-hate) speech. Through ablation studies, we show the relative contributions of annotator embeddings and label text to the model performance, and tested a range of alternative annotator embeddings and label text combinations.

Aiqi Jiang, Arkaitz Zubiaga

The goal of sexism detection is to mitigate negative online content targeting certain gender groups of people. However, the limited availability of labeled sexism-related datasets makes it problematic to identify online sexism for low-resource languages. In this paper, we address the task of automatic sexism detection in social media for one low-resource language -- Chinese. Rather than collecting new sexism data or building cross-lingual transfer learning models, we develop a cross-lingual domain-aware semantic specialisation system in order to make the most of existing data. Semantic specialisation is a technique for retrofitting pre-trained distributional word vectors by integrating external linguistic knowledge (such as lexico-semantic relations) into the specialised feature space. To do this, we leverage semantic resources for sexism from a high-resource language (English) to specialise pre-trained word vectors in the target language (Chinese) to inject domain knowledge. We demonstrate the benefit of our sexist word embeddings (SexWEs) specialised by our framework via intrinsic evaluation of word similarity and extrinsic evaluation of sexism detection. Compared with other specialisation approaches and Chinese baseline word vectors, our SexWEs shows an average score improvement of 0.033 and 0.064 in both intrinsic and extrinsic evaluations, respectively. The ablative results and visualisation of SexWEs also prove the effectiveness of our framework on retrofitting word vectors in low-resource languages.

Tanvi Dinkar, Aiqi Jiang, Gavin Abercrombie et al.

Social media has exacerbated the promotion of Western beauty norms, leading to negative self-image, particularly in women and girls, and causing harm such as body dysmorphia. Increasingly content on the internet has been artificially generated, leading to concerns that these norms are being exaggerated. The aim of this work is to study how generative AI models may encode 'beauty' and erase 'ugliness', and discuss the implications of this for society. To investigate these aims, we create two image generation pipelines: a text-to-image model and a text-to-language model-to image model. We develop a structured beauty taxonomy which we use to prompt three language models (LMs) and two text-to-image models to cumulatively generate 5984 images using our two pipelines. We then recruit women and non-binary social media users to evaluate 1200 of the images through a Likert-scale within-subjects study. Participants show high agreement in their ratings. Our results show that 86.5% of generated images depicted people with lighter skin tones, 22% contained explicit content despite Safe for Work (SFW) training, and 74% were rated as being in a younger age demographic. In particular, the images of non-binary individuals were rated as both younger and more hypersexualised, indicating troubling intersectional effects. Notably, prompts encoded with 'negative' or 'ugly' beauty traits (such as "a wide nose") consistently produced higher Not SFW (NSFW) ratings regardless of gender. This work sheds light on the pervasive demographic biases related to beauty standards present in generative AI models -- biases that are actively perpetuated by model developers, such as via negative prompting. We conclude by discussing the implications of this on society, which include pollution of the data streams and active erasure of features that do not fall inside the stereotype of what is considered beautiful by developers.

Aiqi Jiang, Arkaitz Zubiaga

The growing prevalence and rapid evolution of offensive language in social media amplify the complexities of detection, particularly highlighting the challenges in identifying such content across diverse languages. This survey presents a systematic and comprehensive exploration of Cross-Lingual Transfer Learning (CLTL) techniques in offensive language detection in social media. Our study stands as the first holistic overview to focus exclusively on the cross-lingual scenario in this domain. We analyse 67 relevant papers and categorise these studies across various dimensions, including the characteristics of multilingual datasets used, the cross-lingual resources employed, and the specific CLTL strategies implemented. According to "what to transfer", we also summarise three main CLTL transfer approaches: instance, feature, and parameter transfer. Additionally, we shed light on the current challenges and future research opportunities in this field. Furthermore, we have made our survey resources available online, including two comprehensive tables that provide accessible references to the multilingual datasets and CLTL methods used in the reviewed literature.

Raneem Alharthi, Rajwa Alharthi, Ravi Shekhar et al.

Despite the growing body of research tackling offensive language in social media, this research is predominantly reactive, determining if content already posted in social media is abusive. There is a gap in predictive approaches, which we address in our study by enabling to predict the volume of abusive replies a tweet will receive after being posted. We formulate the problem from the perspective of a social media user asking: ``if I post a certain message on social media, is it possible to predict the volume of abusive replies it might receive?'' We look at four types of features, namely text, text metadata, tweet metadata, and account features, which also help us understand the extent to which the user or the content helps predict the number of abusive replies. This, in turn, helps us develop a model to support social media users in finding the best way to post content. One of our objectives is also to determine the extent to which the volume of abusive replies that a tweet will get are motivated by the content of the tweet or by the identity of the user posting it. Our study finds that one can build a model that performs competitively by developing a comprehensive set of features derived from the content of the message that is going to be posted. In addition, our study suggests that features derived from the user's identity do not impact model performance, hence suggesting that it is especially the content of a post that triggers abusive replies rather than who the user is.

Raneem Alharthi, Rajwa Alharthi, Aiqi Jiang et al.

Abusive language detection has become an increasingly important task as a means to tackle this type of harmful content in social media. There has been a substantial body of research developing models for determining if a social media post is abusive or not; however, this research has primarily focused on exploiting social media posts individually, overlooking additional context that can be derived from surrounding posts. In this study, we look at conversational exchanges, where a user replies to an earlier post by another user (the parent tweet). We ask: does leveraging context from the parent tweet help determine if a reply post is abusive or not, and what are the features that contribute the most? We study a range of content-based and account-based features derived from the context, and compare this to the more widely studied approach of only looking at the features from the reply tweet. For a more generalizable study, we test four different classification models on a dataset made of conversational exchanges (parent-reply tweet pairs) with replies labeled as abusive or not. Our experiments show that incorporating contextual features leads to substantial improvements compared to the use of features derived from the reply tweet only, confirming the importance of leveraging context. We observe that, among the features under study, it is especially the content-based features (what is being posted) that contribute to the classification performance rather than account-based features (who is posting it). While using content-based features, it is best to combine a range of different features to ensure improved performance over being more selective and using fewer features. Our study provides insights into the development of contextualized abusive language detection models in realistic settings involving conversations.

Tanvi Dinkar, Aiqi Jiang, Simona Frenda et al.

Counterspeech, i.e. the practice of responding to online hate speech, has gained traction in NLP as a promising intervention. While early work emphasised collaboration with non-governmental organisation stakeholders, recent research trends have shifted toward automated pipelines that reuse a small set of legacy datasets, often without input from affected communities. This paper presents a systematic review of 74 NLP studies on counterspeech, analysing the extent to which stakeholder participation influences dataset creation, model development, and evaluation. To complement this analysis, we conducted a participatory case study with five NGOs specialising in online Gender-Based Violence (oGBV), identifying stakeholder-informed practices for counterspeech generation. Our findings reveal a growing disconnect between current NLP research and the needs of communities most impacted by toxic online content. We conclude with concrete recommendations for re-centring stakeholder expertise in counterspeech research.

Aiqi Jiang, Arkaitz Zubiaga

Most hate speech detection research focuses on a single language, generally English, which limits their generalisability to other languages. In this paper we investigate the cross-lingual hate speech detection task, tackling the problem by adapting the hate speech resources from one language to another. We propose a cross-lingual capsule network learning model coupled with extra domain-specific lexical semantics for hate speech (CCNL-Ex). Our model achieves state-of-the-art performance on benchmark datasets from AMI@Evalita2018 and AMI@Ibereval2018 involving three languages: English, Spanish and Italian, outperforming state-of-the-art baselines on all six language pairs.

Aiqi Jiang, Xiaohan Yang, Yang Liu et al.

Online sexism has become an increasing concern in social media platforms as it has affected the healthy development of the Internet and can have negative effects in society. While research in the sexism detection domain is growing, most of this research focuses on English as the language and on Twitter as the platform. Our objective here is to broaden the scope of this research by considering the Chinese language on Sina Weibo. We propose the first Chinese sexism dataset -- Sina Weibo Sexism Review (SWSR) dataset --, as well as a large Chinese lexicon SexHateLex made of abusive and gender-related terms. We introduce our data collection and annotation process, and provide an exploratory analysis of the dataset characteristics to validate its quality and to show how sexism is manifested in Chinese. The SWSR dataset provides labels at different levels of granularity including (i) sexism or non-sexism, (ii) sexism category and (iii) target type, which can be exploited, among others, for building computational methods to identify and investigate finer-grained gender-related abusive language. We conduct experiments for the three sexism classification tasks making use of state-of-the-art machine learning models. Our results show competitive performance, providing a benchmark for sexism detection in the Chinese language, as well as an error analysis highlighting open challenges needing more research in Chinese NLP. The SWSR dataset and SexHateLex lexicon are publicly available.

Aiqi Jiang, Arkaitz Zubiaga

Online review platforms are a popular way for users to post reviews by expressing their opinions towards a product or service, as well as they are valuable for other users and companies to find out the overall opinions of customers. These reviews tend to be accompanied by a rating, where the star rating has become the most common approach for users to give their feedback in a quantitative way, generally as a likert scale of 1-5 stars. In other social media platforms like Facebook or Twitter, an automated review rating prediction system can be useful to determine the rating that a user would have given to the product or service. Existing work on review rating prediction focuses on specific domains, such as restaurants or hotels. This, however, ignores the fact that some review domains which are less frequently rated, such as dentists, lack sufficient data to build a reliable prediction model. In this paper, we experiment on 12 datasets pertaining to 12 different review domains of varying level of popularity to assess the performance of predictions across different domains. We introduce a model that leverages aspect phrase embeddings extracted from the reviews, which enables the development of both in-domain and cross-domain review rating prediction systems. Our experiments show that both of our review rating prediction systems outperform all other baselines. The cross-domain review rating prediction system is particularly significant for the least popular review domains, where leveraging training data from other domains leads to remarkable improvements in performance. The in-domain review rating prediction system is instead more suitable for popular review domains, provided that a model built from training data pertaining to the target domain is more suitable when this data is abundant.

Arkaitz Zubiaga, Aiqi Jiang

The unmoderated nature of social media enables the diffusion of hoaxes, which in turn jeopardises the credibility of information gathered from social media platforms. Existing research on automated detection of hoaxes has the limitation of using relatively small datasets, owing to the difficulty of getting labelled data. This in turn has limited research exploring early detection of hoaxes as well as exploring other factors such as the effect of the size of the training data or the use of sliding windows. To mitigate this problem, we introduce a semi-automated method that leverages the Wikidata knowledge base to build large-scale datasets for veracity classification, focusing on celebrity death reports. This enables us to create a dataset with 4,007 reports including over 13 million tweets, 15% of which are fake. Experiments using class-specific representations of word embeddings show that we can achieve F1 scores nearing 72% within 10 minutes of the first tweet being posted when we expand the size of the training data following our semi-automated means. Our dataset represents a realistic scenario with a real distribution of true, commemorative and false stories, which we release for further use as a benchmark in future research.