The Evolution of Search: Three Computing Paradigms

Search is probably the most common activity that humans conduct all the time. A search target can be a concrete item (with a yes or no answer and location information), an abstract concept (such as the most important information on the Web about Xindong Wu), or a plan/path for a specific target with an objective function (like flight scheduling with a minimal travel time), among others. In this article, we propose a Universal Connection Theorem (UCT) to suggest that all physical objects/items in the universe are connected through explicit or implicit relationships. Search is to explore the relationships, using different computing methods, to retrieve relevant objects. Under the UCT theorem, we summarize mainstream search approaches into two categories from the user perspective, deterministic search vs. abstract search, and further distinguish them into three computing paradigms: planning based search, data driven search, and knowledge enhanced search. The planning based paradigm explores search as a planning process in a large search space, by graph traversing with heuristic principles to locate optimal solutions. The data driven paradigm seeks to find objects matching the user's query from a large data repository. Indexing, hashing, information retrieval, and recommendations are typical strategies to tackle the data volumes and select the best answers for users’ queries. The knowledge enhanced search does not aim to find matching objects, but to discover and then meet user's search requirements through knowledge mining. The evolution of these three search paradigms, from planning to data engineering and knowledge engineering, provides increasing levels of challenges and opportunities. This article elaborates the respective principles of these paradigms.