In Linguistics, an anaphor is an expression whose interpretation depends upon another expression in context, namely an antecedent expression. Anaphora resolution is a task of identifying the anaphorical relation between the anaphor and its antecedent. Anaphora resolution is used in many high-level tasks of Natural Language Processing. Traditionally, the rule-based approaches to anaphora resolution rely on the syntactic structures and discourse features. In my study, I implement two semantic approaches on biomedical texts, ontology-dependent method and ontology-independent vector semantic method. The ontology-dependent method will be used to locate the antecedent for noun phrases with determiners while the ontology-independent method will be implemented on pronouns. The results show that the semantic approaches are promising directions in investigating resolutions for anaphora problems in the future.