The ability to understand whether an event requires physical transportation is a crucial step in developing intelligent systems for route planning and mobility optimization. In this study, we explore the automatic classification of calendar events to determine their transport requirements based solely on their textual descriptions. Starting from a small, real-world dataset of labeled events, we construct a diverse semi-synthetic corpus designed to reflect a wide range of workplace and logistical contexts. The dataset includes short, real-life event descriptions annotated with binary transport labels, and is enriched with synthetic examples that preserve linguistic variability and semantic plausibility. We evaluate multiple machine learning approaches, including Decision Trees, Random Forests, and a TF-IDF-based MLP, against a Transformer-based model, DistilBERT, fine-tuned for binary classification. Our results show that contextual language models significantly outperform traditional approaches, achieving over 94% accuracy and a 0.93 F1-score when trained on the augmented dataset. This research contributes a reusable dataset generation strategy and demonstrates the value of language models in understanding implicit mobility needs, serving as a foundation for downstream applications such as real-time route recommendation and automated agenda management.