Query by example for geographic entity search with implicit negative feedback

We propose a method of searching for geographic entities in an unknown place with a query by example in a known place. Geographic entity searches are installed in many Web sites such as those for shopping and restaurants. Most of the sites hold classic attribute-based or keyword-based interfaces for entity retrieval; however, specifying each attribute users want is time-consuming, and keywords are not effective for representing users’ complex intentions. The proposed query by example method in a map interface allows users to intuitively query by selecting entities in places they know well. The most similar entities to an input are returned based on the similarity varying with individuals. Our proposed method is robust for estimating the similarity using not only selected examples, but also implicit negative feedback, which is predicted by how the user selects examples as a query in the map interface. Experimental results proved the effectiveness of our method, and the performance exceeded that of a previously proposed method.