Fast nearest Neighbor Browsing & Search with Keywords
|Name||Fast nearest Neighbor Browsing & Search with Keywords|
Conventional abstraction queries, like vary search and nearest neighbor retrieval, involve solely conditions on objects’ geometric properties. Today, several trendy applications involve novel kinds of queries that aim to seek out objects satisfying each a abstraction predicate, and a predicate on their associated texts. as an example, rather than considering all the restaurants, a nearest neighbor question would instead elicit the eating house that's the highest among those whose menus contain “steak, spaghetti, brandy” all at an equivalent time. Presently the most effective resolution to such queries is predicated on the IR2-tree, which, as shown during this paper, features a few deficiencies that seriously impact its potency. Impelled by this, we tend to develop a replacement access methodology known as the abstraction inverted index that extends the standard inverted index to address flat knowledge, and comes with algorithms that may answer nearest neighbor queries with keywords in real time. As verified by experiments, the projected techniques outgo the IR2-tree in question latent period considerably, typically by an element of orders of magnitude.
|ieee paper year||2014|