Weaving Text into Tables

Abstract

In this paper, we showcase JIGSAW, a system that is able to shape unstructured text into structured tables for user-defined schemas. In short, to structure text into tables, JIGSAW leverages the lexico-syntactic structure imposed by linguistic annotations (e.g., part-of-speech, named entities, temporal and numerical expressions) on natural language text. We describe how challenging knowledge-centric tasks such as question answering, summarization, and analytics can be greatly simplified with the help of JIGSAW.