Output-Sensitive Autocompletion Search

Abstract

Abstract. We consider the following autocompletion search scenario: imagine a user of a search engine typing a query; then with every keystroke display those completions of the last query word that would lead to the best hits, and also display the best such hits. The following problem is at the core of this feature: for a fixed document collection, given a set D of documents, and an alphabetical range W of words, compute the set of all word-in-document pairs (w, d) from the collection such that w ∈ W and d ∈ D. We present a new data structure with the help of which such autocompletion queries can be processed, on the average, in time linear in the input plus output size, independent of the size of the underlying document collection. At the same time, our data structure uses no more space than an inverted index. Actual query processing times on a large test collection correlate almost perfectly with our theoretical bound.

Location-Based Instant Search

“…interactive search, auto-complete search, or type-ahead search) has become in many search systems. Bast et al [3,2] proposed indexing techniques for efficient auto-complete search. The studies in [13,6] investigated how to do error-tolerant interactive search.…”

“…(1) All the prefixes represented by a radix tree node have the same interval. For instance, prefixes C and Ch, represented by the same node in Figure 5, have the same interval [1,2]. Therefore, we can use an interval to represent a radix tree node to be stored in the prefix filter, instead of storing all the prefixes that share the same interval.…”

“…Figure 7 shows the prefix filters with intervals on node p in the FEH prefix index in our running example. For instance, the record-selective prefix Ch is added to the O-filter on node p. It is stored as its corresponding interval [1,2], with the list of the records that have the prefix (record 7). The intervals added to a filter on the node p form a radix forest (a set of disjoint radix trees), shown on the left side of the filter in the figure.…”

“…For example, Figure 8(a) shows the radix forest of record-selective prefixes. [1,1] and [2,2] are removed from the compressed O-filter, as they extend another interval [1,2] in the same filter. The O-filter is physically stored as in Figure 8(b).…”

“…The prefix filter on the R*-tree node p. query interval from the C-filter, and visit the consistent R*-tree children. For instance, to answer the query with the prefix Ch, we lookup its corresponding interval [1,2] from the global vocabulary. When visiting the R*-tree node p, we lookup [1,2] from the O-filter, and then directly retrieve the records without visiting any other children nodes of p.…”