Automation of Corporate Language Services: Hybrid AI Approaches to Text Anonymization and Termbase Expansion

This thesis examines application scenarios of Artificial Intelligence (AI) and Automation to the workflows of a corporate Language Services (LS) department through two practical case studies. It focuses on hybrid AI approaches to Text Anonymization and Termbase (TB) Expansion, combining theoretical analysis with applied system design and evaluation. The first case study presents an LLM-based anonymization tool developed to ensure compliance with data protection regulations that integrates with the already existing platforms and workflows. The second case study proposes a semi-automated pipeline that performs subdomain identification, term extraction, concept consolidation, metadata augmentation and relation identification to enhance the internal TB. Both solutions combine deterministic processes, probabilistic models and human expertise, demonstrating how these components mitigate one another’s limitations. Results show that AI-assisted workflows can reduce manual effort, improve scalability and support consistency, while expert human validation remains essential to ensure quality and compliance. The thesis situates these findings within the broader technological, professional and ethical transformation of the language industry, concluding that hybrid AI paradigms could be a way to enable the effective and responsible deployment of AI in corporate environments.