Output-sensitive autocompletion search

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 wordin-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.

Practical rearrangement methods for dynamic double‐array dictionaries

“…If a key is registered without TAIL, the associated value is embedded in the corresponding leaf BASE element, eg, BASE [4] = 0. Embedding is the most practical dictionary implementation using dynamic DA tries.…”

“…Then, a new leaf node labeled a is defined and linked to TAIL [5]. Consequently, TAIL [4] becomes empty. At the same time, another new leaf node labeled c is also defined.…”

A Network-Aware Approach for Searching As-You-Type in Social Media

“…Main algorithmic components. When accessing the CT-IL index, inverted list entries are consumed in some IL(t) only when the items they refer to are candidates (they appear in at least one D t buffer, which may not necessarily be Dt itself) 2 . We keep in lists called CILt (for consumed IL entries) the items read (hence known candidates) in the inverted lists (virtual or concrete), for t being either in {t1, .…”

“…unseen_users(i, t) then reflects at any moment during the run of the algorithm the difference between the number of taggers of i with t already visited and one of either During the algorithm's run, for known candidates i of some Dt, we accumulate in sf (i | s, t) the social score (initially set to 0). 2 The rationale is that our algorithm does not make any "wild guesses", avoiding reads that may prove to be irrelevant and thus leading to sub-optimal performance. Each time we visit a user u having a triple (u, i, t) in her p-space (Algorithm 2), we can 1. update sf (i | s, t) by adding σ + (s, u) to it, and 2. decrement unseen_users(i, t); when this value reaches 0, the social frequency value sf (i | s, t) is final.…”

“…, tr−1}, (d) reinitialize the network exploration (the queue H) to start from the seeker, in order to visit again p-spaces looking for triples for the new prefix, tr. (This amounts to the following changes in Algorithm 1: among its initialisation steps (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12), the steps (4-8) are removed, and new steps for points (a) and (c) above are added. )…”