Extracting Enterprise Vocabulary Using Linked Open Data

A common vocabulary is vital to smooth business operation, yet codifying and maintaining an enterprise vocabulary is an arduous, manual task. We present a fully automated process for creating an enterprise vocabulary, by extracting terms from a domain-specific corpus, and extracting their types from LOD (Linked Open Data). We applied this process to create a vocabulary for the IT industry, using 58 Gartner analyst reports as a corpus, and the LOD subset consisting of DBpedia and Freebase. We present novel techniques for linking, cleansing, and extending the types in this LOD subset, resulting in an improvement of 55% for our IT domain results. We further improved our results through NER over the corpus. Our NER training is completely automated, exploiting Wikipedia and DBpedia. Altogether, we achieved 46.3% recall and 78.1% precision.