DataStory: Intentional database querying to facilitate interactive data storytelling

The aim of the DATASTORY project is to revise the definition of what a query is (currently: a declarative expression over a set of tables producing a set of tuples), both with respect to its specification, and with respect to its produced result. The new era of query specification that we envision, replaces the declarative specification of SQL, Datalog, rel. calculus with an algebra of intentional operators that allow the user to assess the situation presented by a subset of a data space via comparisons of the retrieved values to similar ones, explain and analyze the demonstrated situation with more detailed data, predict future values etc. The new era of query answering envisions replacing the current recordset-as-an-answer query result with a data story, i.e., a report that complements the traditional query answer with results of auxiliary queries that contextualize and explain the answer as well as with concise and meaningful highlights and commentaries, packaged with appropriate visualization and textual descriptions that ultimately enhance user insights.

To facilitate the production of data stories, we envision that the query answering subsystem of a database engine is complemented with a data narration engine. The data narration engine relies on the rigorous semantics of in the intentional operators to produce auxiliary queries and machine learning tasks for the answering of a query. An optimizer takes advantage of multi-query optimization opportunities, pre-existing machine learning models and cached results to efficiently execute the respective tasks. A highlight selector assesses the significance of the produced results on the basis of an interestingness model, prunes the non-interesting ones and passes highlighted data and models to a story production module that automatically generates textual commentaries and visualizations and assembles them into a data story. The analyst can interactively refine and enlarge the story with complementary tasks.