Semi-supervised learning methods allow to use a priori knowledge to guide the classification algorithm in group discovery. In this work, we propose a new hierarchical agglomerative clustering algorithm (HAC) that takes into account two sources of information associated with the same objects. This method, called compromise HAC (hclustcompro), allows a compromise between the hierarchies obtained from each source taken separately. A convex combination of the dissimilarities associated with each of the sources is used to modify the dissimilarity measure in the classical HAC algorithm. The choice of the mixing parameter is the key point of the method. We propose an objective function to be minimized based on the absolute difference of correlations between initial dissimilarities and cophenetics distances, as well as a resampling procedure to ensure the robustness of the choice of the mixing parameter. We illustrate our method with archaeological data from the Angkor site in Cambodia.