Analyzing the Use of Large Language Models for Content Moderation with ChatGPT Examples

Franco, Mirko; Gaggi, Ombretta; Palazzi, Claudio E.

doi:10.1145/3599696.3612895

Cited by 8 publications

(3 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[5][6][7][8], text generation (content creation, code generation for programming languages, etc.) [9,10], question answering, conversational AI [11,12], language summarization [13,14], content recommendation and moderation [15,16], search engines, web pages and documents ranking [17], and data extraction and knowledge graph creation [18,19].…”

Section: State-of-the-artmentioning

confidence: 99%

esCorpius-m: A Massive Multilingual Crawling Corpus with a Focus on Spanish

Gutiérrez-Fandiño,

Pérez-Fernández,

Armengol-Estapé

et al. 2023

Applied Sciences

View full text Add to dashboard Cite

In recent years, transformer-based models have played a significant role in advancing language modeling for natural language processing. However, they require substantial amounts of data and there is a shortage of high-quality non-English corpora. Some recent initiatives have introduced multilingual datasets obtained through web crawling. However, there are notable limitations in the results for some languages, including Spanish. These datasets are either smaller compared to other languages or suffer from lower quality due to insufficient cleaning and deduplication. In this paper, we present esCorpius-m, a multilingual corpus extracted from around 1 petabyte of Common Crawl data. It is the most extensive corpus for some languages with such a level of high-quality content extraction, cleanliness, and deduplication. Our data curation process involves an efficient cleaning pipeline and various deduplication methods that maintain the integrity of document and paragraph boundaries. We also ensure compliance with EU regulations by retaining both the source web page URL and the WARC shared origin URL.

show abstract

Section: State-of-the-artmentioning

confidence: 99%

esCorpius-m: A Massive Multilingual Crawling Corpus with a Focus on Spanish

Gutiérrez-Fandiño,

Pérez-Fernández,

Armengol-Estapé

et al. 2023

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…The high interconnectedness of these online communities, both within and across different social media platforms, can hence lead to the swift proliferation of extreme viewpoints and associated mis/disinformation. Social networks like Facebook employ a combination of automated systems and human moderators to identify and combat misinformation on their platforms [32], [33]. Machine learning algorithms are used to detect potentially problematic content, which is then reviewed by human moderators to determine whether it violates the platform's community standards [33].…”

Section: Introductionmentioning

confidence: 99%

“…Machine learning algorithms are used to detect potentially problematic content, which is then reviewed by human moderators to determine whether it violates the platform's community standards [33]. However, this approach faces several challenges, such as the risk of biased decisions by automated systems and the difficulty of balancing free speech with the need to protect users from harm [32], [33]. Additionally, the increasing sophistication of large language models (LLMs) like ChatGPT has raised questions about their potential use in content moderation, as well as the associated risks and limitations [32].…”

Section: Introductionmentioning

confidence: 99%

Softening Online Extremes Using Network Engineering

Restrepo,

Moreno,

Illari

et al. 2024

IEEE Access

View full text Add to dashboard Cite

The prevalence of dangerous misinformation and extreme views online has intensified since the onset of Israel-Hamas war on 7 October 2023. Social media platforms have long grappled with the challenge of providing effective mitigation schemes that can scale to the 5 billion-strong online population. Here, we introduce a novel solution grounded in online network engineering and demonstrate its potential through small pilot studies. We begin by outlining the characteristics of the online social network infrastructure that have rendered previous approaches to mitigating extremes ineffective. We then present our new online engineering scheme and explain how it circumvents these issues. The efficacy of this scheme is demonstrated through a pilot empirical study, which reveals that automatically assembling groups of users online with diverse opinions, guided by a map of the online social media infrastructure, and facilitating their anonymous interactions, can lead to a softening of extreme views. We then employ computer simulations to explore the potential for implementing this scheme online at scale and in an automated manner, without necessitating the contentious removal of specific communities, imposing censorship, relying on preventative messaging, or requiring consensus within the online groups. These pilot studies provide preliminary insights into the effectiveness and feasibility of this approach in online social media settings.

show abstract