Equity is an end-to-end system for canonicalizing mentions in ad-hoc tables and their surrounding contexts, by linking mentions of entities, classes, concepts and quantities to a knowledge base. Equity employs an efficient algorithm and achieves an overall linking precision of about 92%, and 93% for quantity mentions. The Equity system is a first building block in our longer-term research towards making sense of quantities in tables and text.
We propose a distantly supervised Markov Random Fields (MRF) model for canonicalizing ad-hoc tables, handling classes, concepts, entities and quantities in both table cells and surrounding text. We present an efficient algorithm, based on random walks, for computing high-quality solutions. For more details refer to our paper.
- Annotated Tables [ JSON ]