Some points of interest could fall under multiple point of interest types; for example, a hotel could also have a restaurant operating under the same name. In such case only the primary type is considered. For real-life application, we would probably want to include all classifications for each resource and let the search algorithm to determine the optimal order. For an itinerary website, the benefit of showing a potential match probably outweigh the cost of serving an irrelevant result. Machine Learning algorithms will help prioritize the results and improve accuracy and specificity over time.