Text-To-Query: dynamically building structured analytics to illustrate textual content

摘要

Successfully structuring information in databases, OLAP cubes, and XML is a crucial element in managing data nowadays. However this process brought new challenges to usability. It is difficult for users to switch from common communication means using natural language to data models (e.g., database schemas) that are hard to work with and understand, especially for occasional users. This important issue is under intense scrutiny in the database community (e.g., keyword search over databases and query relaxation techniques), and the information extraction community (e.g., linking structured and unstructured data). However, there is still no comprehensive solution that automatically generates an OLAP (Online Analytical Processing) query and chooses a visualization based on textual content with high precision. We present such a method. We discuss how to dynamically generate interpretations of a textual content as an OLAP query, select the best visualization, and retrieve on the fly corresponding data from a data warehouse. To provide the most relevant aggregation results, we consider the user's actual context, described by a document's content. Moreover we provide a prototypical implementation of our method, the Text-To-Query system (T2Q) and show how T2Q can be successfully applied to an enterprise scenario as an extension for an office application.