Centre for Cyberhate Research & Policy: Real-Time Scalable Methods & Infrastructure for Modelling the Spread of Cyberhate on Social Media

网络仇恨研究中心

• 批准号: ES/P010695/1
• 负责人: Matthew Williams
• 金额: $ 234.35万
• 依托单位: CARDIFF UNIVERSITY
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2017
• 资助国家: 英国
• 起止时间: 2017 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=ES%2FP010695%2F1
• 关键词: Centre; Cyberhate; Research; Policy; Real

The UK Government's Hate Crime Action Plan (Home Office 2016) stresses the need to tackle hate speech on social media by bringing together policymakers with academics to improve the analysis and understanding of the patterns and drivers of cyberhate and how these can be addressed. Furthermore, the recent Home Affairs Select Committee Inquiry (2016) 'Hate Crime and its Violent Consequences' highlighted the role of social media in the propagation of hate speech (on which the proposers were invited to provide evidence). This proposal acknowledges the migration of hate to social media is non-trivial, and that empirically we know very little about the utility of Web based forms data for measuring online hate speech and counter hate speech at scale and in real-time. This became particularly apparent following the referendum on the UK's future in the European Union, where an inability to classify and monitor hate speech and counter speech on social media in near-real-time and at scale hindered the use of these new forms of data in policy decision making in the area of hate crime. It was months later that small-scale grey literature emerged providing a 'snap-shot' of the problem (Awan & Zempi 2016, Miller et al. 2016). In partnership with the UK Head of the Cross-Government Hate Crime Programme at the Department for Communities and Local Government (DCLG), and the London Mayor's Office for Policing and Crime's (MOPAC) new Online Hate Crime Hub, the proposed project will co-produce evidence on how social media data, harnessed by new Social Data Science methods and scalable infrastructure, can inform policy decision making. We will achieve this by taking the social media reaction to the referendum on the UK's future in the European Union as a demonstration study, and will co-develop with the Policy CI transformational New Forms of Data Capability contributions including: (i) semi-automated methods that monitor the production and spread of cyberhate around the case study and beyond; (ii) complementary methods to study and test the effectiveness of counter speech in reducing the propagation of cyberhate, and (iii) a technical system that can support real time analysis of hate and counter speech on social media at scale following 'trigger events', integrated into existing policy evidence-based decision-making processes. The system, by estimating the propagation of cyberhate interactions within social media using machine learning techniques and statistical models, will assist policymakers in identifying areas that require policy attention and better targeted interventions in the field of online hate and antagonistic content.

英国政府的仇恨犯罪行动计划（内政部2016年）强调，需要通过将政策制定者与学者聚集在一起，以改善对网络仇恨模式和驱动因素的分析和理解，以及如何解决这些问题，来解决社交媒体上的仇恨言论。此外，最近的内政事务特别委员会调查（2016年）“仇恨犯罪及其暴力后果”强调了社交媒体在传播仇恨言论中的作用（提议者被邀请提供证据）。该提案承认仇恨迁移到社交媒体是不平凡的，并且根据经验，我们对基于Web的表格数据用于测量在线仇恨言论和大规模实时反击仇恨言论的实用性知之甚少。这一点在关于英国在欧盟的未来的全民公决之后变得尤为明显，在全民公决中，无法对社交媒体上的仇恨言论和反仇恨言论进行近实时和大规模的分类和监测，阻碍了在仇恨犯罪领域的政策决策中使用这些新形式的数据。几个月后，出现了小规模的灰色文献，提供了问题的“快照”（Awan & Zempi 2016，米勒et al. 2016）。该项目与英国社区和地方政府部（DCLG）跨政府仇恨犯罪项目负责人以及伦敦市长警务和犯罪办公室（MOPAC）新的在线仇恨犯罪中心合作，将共同提供证据，说明社交媒体数据如何利用新的社会数据科学方法和可扩展的基础设施，为政策决策提供信息。我们将通过将社交媒体对英国在欧盟未来的公投的反应作为示范研究来实现这一目标，并将与政策CI转型新形式的数据能力贡献共同开发，包括：（i）监控案例研究及其他范围内网络仇恨的生产和传播的半自动化方法; ㈡研究和测试反言论在减少网络仇恨传播方面的有效性的补充方法，㈢能够支持在“触发事件”之后对社交媒体上的仇恨和反言论进行大规模真实的时间分析的技术系统，纳入现有的政策循证决策进程。该系统通过使用机器学习技术和统计模型估计社交媒体中网络仇恨互动的传播，将协助决策者确定需要政策关注的领域，并在网上仇恨和敌对内容领域采取更有针对性的干预措施。

项目成果

The Effect of the Brexit Vote on the Variation in Race and Religious Hate Crimes in England, Wales, Scotland and Northern Ireland

英国脱欧公投对英格兰、威尔士、苏格兰和北爱尔兰种族和宗教仇恨犯罪变化的影响

• DOI: 10.1093/bjc/azac071
• 发表时间: 2023
• 期刊: The British Journal of Criminology
• 作者: Williams M

Antisemitism on Twitter: Collective Efficacy and the Role of Community Organisations in Challenging Online Hate Speech

• DOI: 10.1177/2056305120916850
• 发表时间: 2020-04-01
• 期刊: SOCIAL MEDIA + SOCIETY
• 影响因子: 5.2
• 作者: Ozalp, Sefa;Williams, Matthew L.;Mostafa, Mohamed
• 通讯作者: Mostafa, Mohamed

A Fuzzy Approach to Text Classification With Two-Stage Training for Ambiguous Instances

• DOI: 10.1109/tcss.2019.2892037
• 发表时间: 2019-03
• 期刊: IEEE Transactions on Computational Social Systems
• 影响因子: 5
• 作者: Han Liu;P. Burnap;Wafa Alorainy;M. Williams

Fuzzy Multi-task Learning for Hate Speech Type Identification

• DOI: 10.1145/3308558.3313546
• 发表时间: 2019-05
• 期刊: The World Wide Web Conference
• 作者: Han Liu;P. Burnap;Wafa Alorainy;M. Williams

Scmhl5 at TRAC-2 Shared Task on Aggression Identification: Bert Based Ensemble Learning Approach

• 发表时间: 2020-05
• 作者: Han Liu;P. Burnap;Wafa Alorainy;M. Williams