How to Build a Controlled Vocabulary of Concepts for Deep Semantic Search

Semantic Search (SS) is well understood and widely applied in search engines, WordNets (PWN) and ontologies. It is based on Semantic Relations (SR) between different concepts. SS in general does not take account of Intra-Concept Relations (ICR). This article describes how concepts can be formally composed as Semantic Compounds (SC) of sub concepts and Semantic Primes (SP) using ICRs. SMs are stored in a Controlled Vocabulary of Concepts (CVC) which is the fundamental for the implementation of a novel Deep Semantic Search (DSS) algorithm also called Search by Meaning (SbM). The evaluation of SbM has been conducted with a large energy ontology (EnArgus) and the mobile app rObby for the automated notification on scientific publications. SbM finds more relevant concepts and publications compared to traditional SS. The results of SbM are also evaluated against those of Google Search (GS) and Princeton WordNet (PWN).