Assessing Switzerland’s reputation using fine-tuned RAG-enhanced LLMs

Supervisors: Narek Andreasyan, Edy Portmann

Contact person: Narek Andreasyan

Student: Sipan Hoj

Project status: Ongoing

Year: 2025

Project Description

This Master’s thesis offers the opportunity to contribute to a larger research project that investigates the reputation of Switzerland in times of artificial intelligence (AI). The research focuses on the analysis of multilingual and multimodal media content that influences the reputation of Switzerland in international contexts. By employing Large Language Models (LLMs) for automated content analysis, the project aims to measure and analyze how Switzerland is portrayed and its reputation shaped across media platforms.

Research Question

  • How can AI be employed to quantitatively measure Switzerland’s country reputation across various media sources (news websites, social media, etc.)?

Research Focus

  • Development and implementation of a system based on LLMs and Retrieval-Augmented Generation (RAG) to analyze Switzerland's representation in various media sources.
  • The system shall be optimized for processing large datasets referring to multilingual (English, German, French, Italian, etc.) and multimodal (text, images, audio, video) media content in order to gain insights into reputation drivers and sentiments relating to Switzerland’s country reputation.

Methods and Tools

  • Media Monitoring Data Access to data from a media monitoring company tracking media representations of Switzerland.
  • AI Models Implementation of open-source (or proprietary) AI models, which can be customized and fine-tuned for the research goals.
  • Multimodal Analysis Processing and analyzing a combination of data fromats including text, image, audio and video relevant to Switzerland’s reputation in the international public sphere.

Qualifications

  • Master's student in Computer Science or related field with a strong understanding of machine learning and natural language processing (NLP) and experience in AI applications.
  • Excellent written and verbal communication skills in German and English
  • A proactive, research-oriented mindset, eager to contribute to a project that intersects computer science and social sciences.

Keywords: Large Language Modles (LLMs); Retrieval-Augmented Generation (RAG); Fine-tuning; Reputation; Swiss Reputation assessment; LLM-based Reputation Analysis

Document: Not yet available