• BERT
• DID
• ESCO
• Fake News
• False News
• GPT
• LLM
• Match Skills
• Misinformation
• NER
• NLP
• Ontology
• Self-Sovereign Identity
• Skill Extraction
• Social Content
• Social Network
• Verifiable Credential

The thesis focuses on assessing the trustworthiness of content shared on social media using Self-Sovereign Identity (SSI) technology. This technology allows users to provide digital certificates, called Verifiable Credentials (VCs) issued to their identity represented by Decentralized Identifiers (DIDs), to prove their authority on specific topics without revealing other personally identifiable information such as real name or surname.

The proposed system, called CAVS (Credibility Assessment and Verification System), aims to reduce the spread of fake news, with a definition given in the thesis, by matching the VC with statement content. For doing so, ESCO ontology is used in CAVS to match the author's skills in VCs with the content they publish using the support of AI tools like the BERT model for skill extraction.

The thesis also explores the integration of CAVS with a decentralized social network called DeSo. It includes a simulation comparing the performance of traditional fact-checkers who know the truthfulness of news with fact-checkers using the CAVS system. The simulations exploit a synthetic dataset generated with GPT3.5 and show that the system, while needing wide adoption, can significantly improve information trustworthiness in a privacy-preserving manner, maintaining freedom of expression.