The first purpose of this project is to explore the semantic differences and similarities in odor vocabularies across multiple languages and domains, employing a data-driven approach that utilizes natural text corpora. In the first part of the project, we focused extensively on the English language, using a wide range of language models, which led to two publication in high-impact journals. In the second part, we focused on differences in odor and flavor vocabularies in three different domains: wine, perfume and food. Also this work resulted in a publication in a high-impact journal. As we move forward, our research will now expand to include additional languages: Swedish, Turkish, Italian and Thai, analyzing the semantic organization of olfactory vocabularies in these languages. The second purpose is to explore how artificial intelligence (AI) systems—specifically large language models like ChatGPT—can learn to understand sensory experiences such as smells, even though they don't have noses or physical senses like humans do. The aim of this project is to build the world’s first "smell-enriched" language model—an AI that understands olfactory (smell-related) experiences. To do this, we will collect detailed descriptions of smells from people, along with ratings of how they perceive different odors. We will also use chemical data that describes the actual molecular properties of different smells. These rich, varied data sources will be combined using a new type of machine learning technique that helps the AI learn from different types of information.