Semantic annotation has become an important piece of information within corpus linguistics. This information is usually included for every lexical unit of the corpus providing a more exhaustive analysis of language. There are some resources such as lexicons or ontologies that allow this type of annotation. However, expanding these resources is a time-consuming task. This paper describes a simple NLP baseline for increasing accuracy of the existing semantic resources of the UCREL Semantic Analysis System (USAS). In our experiments, Spanish token accuracy is improved by up to 30% using this method.