This paper presents a new ontology matching algorithm, OntoPhil. The algorithm relies on the exploitation of some initial correspondences or binding points that connect the two ontologies used as input. First, it computes these binding points using a new lexical similarity measure which combines the information from a terminological matcher and an external one. Next, it discovers new binding points, by taking the initial ones as basis and by exploiting the specific features of the external structure of the ontologies matched. Finally, the binding points are automatically sifted out to obtain the final alignment. The proposed algorithm was tested on the benchmarks provided by the well known evaluation initiative OAEI, and compared to other matching algorithms. The experimental results show that OntoPhil is an effective approach and outperform other algorithms that share the same principles.