Semantic interpretation and integration of open data tables

Abstract

Open data initiatives from governments and corporations publish large amounts of tabular data regularly, that contain useful, actionable knowledge relevant to several stakeholders. However, this knowledge is fragmented across several tables without any overarching schema in which they are organized. Open data portals by themselves are theme agnostic-the data is not organized for any specific purpose or use case. In this research, we address the question of interpreting open data tables and integrating them in a theme-agnostic fashion. Given a collection of tables, we propose algorithms to search the Linked Open Data (LOD) cloud to look for one or more themes that could describe this collection. Then, the columns of the tables are resolved to entities and relationships within the identified themes. A given collection of tables would result in several themes, with semantic associations distributed across different themes. To represent and reason about such a collection we present a knowledge representation framework called Many Worlds on a Frame (MWF), where knowledge is organized within one or more thematic ‘worlds’ each of which in turn relate to one another to form the global knowledge frame. We also generate an ‘Open Knowledge Graph’ using rich N-Quad W3C standard, in line with the thematic outputs allowing for the execution of SPARQL queries against the Open Knowledge Graph by themes or schematic associations. Finally, we demonstrate our model using a practical use case of semantically integrating open data tables from data.gov.in. Here, a set of georeferenced themes, such as places and locations, together with other meaningful concepts from LOD convert the open data tables into a semantic data mesh of interrelated and intuitively traversable data.