Recent advancements in the semantic Web has resulted in the proliferation of a number of domain-specific ontologies and vocabularies. Consequently, the ability to effectively and efficiently search ontologies has emerged as an important problem. More often the users, ranging from naive to domain-experts, who seek ontologies for their application (e.g., analysis of patient data) either use existing search engines such as Swoogle (Ding et al. 2004) or solicit suggestions from peers (e.g., via mailing lists). Unsurprisingly, this is a very cumbersome and labor intensive process. Towards this end, in this paper we propose LexSearch - a technique that facilitates data-driven (semi-) automatic ontology search. Specifically, LexSearch leverages semantic annotations or tags created by the users and experts to generate a triple-based data model which is applied for selecting domain-specific ontologies. The discovered ontologies are then ranked primarily based on content coverage. An important aspect of LexSearch is to provide an uniform ontology-language agnostic framework for finding multiple ontologies that takes into consideration various relationships between the ontological concepts.