Page 190 - tech fest 2025 ב״ש חוברת תקצירים

190
Anonymizing Text with AI: A Secure Solution for Sensitive Information
SE-E-10
Adiv Eliyahu; adiveliyahu3@gmail.com Itay Dosik; itaydosik@gmail.com
Advisor: Dr. Marina Litvak
SCE - Shamoon College of Engineering, Be’er-Sheva
With the growing use of textual data in domains such as healthcare, law, and public administration, protecting sensitive information has become increasingly important. This project explores a method for text anonymization that combines named entity recognition (NER) with a large language model (LLM) to assist in privacy-related classification. The integration aims to identify and mask sensitive entities in text while maintaining overall data utility. The system offers a straightforward interface: Users can input raw text and receive anonymized output efficiently without the need for complex setup or configurations. The approach seeks to contribute toward safer data handling practices by reducing the risk of exposure and retaining as much non-sensitive information as possible.
Keywords: anonymization, large language model, named entity recognition, privacy, sensitive data
Automated Popular Science Article Generation with Text Simplification
SE-E-11
Tomer Netzer; tomer.netzer14@gmail.com Neria Lahyani; neria616@gmail.com Yossi Yosupov; yossiii050@gmail.com
Advisor: Dr. Marina Litvak
SCE - Shamoon College of Engineering, Be’er-Sheva
Scientific publications are essential for knowledge dissemination, but their complexity limits accessibility. As research advances, simplified content becomes increasingly important for informed decisions in fields such as healthcare and technology. This project has developed an large language model (LLM)-based system that simplifies scientific texts while maintaining accuracy and coherence. Using natural language model (NLP) techniques, especially the T5 Transformer model, it processes multilingual texts and adapts to readability. Development involved dataset analysis, model training, and performance evaluation.
The final system features a user-friendly interface for comparing original and simplified texts, bridging the gap between complex research and public understanding. Future enhancements will improve efficiency, expand language support, and address challenges in handling highly technical content.
Keywords: LLM, multilingual Support, NLP, scientific accessibility, text simplification