Knowledge-Based Techniques for Scholarly Data Access: Towards Automatic Curation

Accessing up-to-date and quality scientific literature is a critical preliminary step in any research activity. Identifying relevant scholarly literature for the extents of a given task or application is, however a complex and time consuming activity. Despite the large number of tools developed over the years to support scholars in their literature surveying activity, such as Google Scholar, Microsoft Academic search, and others, the best way to access quality papers remains asking a domain expert who is actively involved in the field and knows research trends and directions. State of the art systems, in fact, either do not allow exploratory search activity, such as identifying the active research directions within a given topic, or do not offer proactive features, such as content recommendation, which are both critical to researchers. To overcome these limitations, we strongly advocate a paradigm shift in the development of scholarly data access tools: moving from traditional information retrieval and filtering tools towards automated agents able to make sense of the textual content of published papers and therefore monitor the state of the art. Building such a system is however a complex task that implies tackling non trivial problems in the fields of Natural Language Processing, Big Data Analysis, User Modelling, and Information Filtering. In this work, we introduce the concept of Automatic Curator System and present its fundamental components.