LOV stands for Linked Open Vocabularies. This name is derived from LOD, standing for Linked Open Data. Let’s assume that the reader is somehow familiar with the latter concept, otherwise a visit to http://linkeddata.org/ or http://www.w3.org/2013/data/ will help to figure it before further reading.
Data on the Web use properties (aka predicates) and classes (aka types) to describe people, places, products, events, and any kind of things whatsoever. In the data “Mary is a person, her family name is Watson, she lives is the city of San Francisco”, “Person” is the class of Mary, “City” is the class of San Francisco, “family name” and “lives is” are properties used to describe a person, the latter acting also as a link between a person and a place.
A vocabulary in LOV gathers definitions of a set of classes and properties (together simply called terms of the vocabulary), useful to describe specific types of things, or things in a given domain or industry, or things at large but for a specific usage.
Terms of vocabularies also provide the links in linked data, in the above case between a Person and a City. The definitions of terms provided by the vocabularies bring clear semantics to descriptions and links, thanks to the formal language they use (some dialect of RDF such as RDFS or OWL). In short, vocabularies provide the semantic glue enabling Data to become meaningful Data.