On the Automation of Similarity Information Maintenance in Flexible Query Answering Systems

This paper proposes a method for automatic maintaining the similarity information for a particular class of Flexible Query Answering Systems (FQAS). The paper describes the three main levels of this approach: the first one deals with learning the distance measure through interaction with the user. Machine-learning techniques, such as reinforcement learning, can be used to achieve this. The second level tries to build a good representation of the learned distance measure. This level uses distance geometry and multidimensional optimization methods. The last level of automation uses statistical optimization techniques to further decrease the dimension of the similarity data.