GnosisMiner: Reading Order Recommendations over Document Collections

Given a document collection, existing systems allow users to locate documents either using search keywords or by navigating through some predefined organization of the collection. Other approaches help the user understand a collection by generating summaries or clusters of the documents at hand. However, often users would like to understand how the documents may be related to each other and access them in some logical order. In this work, we present an interactive reading recommendation system, called GnosisMiner. Given a collection of documents and a theme, the system returns a partial order of documents relevant to that theme organized from more general to more specific. The recommended reading order resembles the human approach of learning as we typically start our path to knowledge from more general documents that help us understand the domain and then we proceed with more specific, more specialized documents to increase our knowledge of the matter.