Automating web Table Columns to Knowledge Base Mapping using Translation Embedding

Extracting semantic information from web tables is a heavily researched area in natural language processing (NLP) and knowledge base augmentation. Identifying the kind of information each table column contains makes it easy for computers to read, understand and use the information in a table. Mapping the columns to concepts from an ontology solves this issue. This paper introduces Col2Pedia, a novel supervised learning based approach to map web table columns to concepts from a knowledge base, automatically. Col2Pedia leverages the behaviour of columns of a table when taken as a pair to infer mappings for these columns with high precision and recall.