Question-answering systems face a challenge related to the process of deciding automatically about the veracity of a given answer. This issue is particularly problematic when handling open-ended questions. In this paper, we propose a multilingual semantic similarity-based approach to estimate the similarity score between the user’s answer and the right one saved in the data tier. This approach is mainly based on semantic information notably the synonymy relationships between words and syntactico-semantic information especially semantic class and thematic role. It supports three languages: English, French and Arabic. Our approach is applied to a multilingual ontology-based question-answering training for Alzheimer’s disease patients. The performance of the pro- posed approach was confirmed through experiments on 20 patients that promising capabilities in identifying literal and some types of intelligent similarity.