This work approaches natural language processing techniques. It highlights theoretical fundamentals needed in order to build syntactic and semantic parse trees that capture the complete meaning of user utterances for search and rescue tasks. For this purpose Lambda Calculus and Combinatory Categorial Grammar formalisms are used. This work also addresses an algorithm that learns latent variables to a Probabilistic Combinatory Categorial Grammar model, as well as its application to derive logical forms for a small set of commands given in Portuguese.