今天发现一篇google的论文,"Hyper-Local,Direction-Based Ranking of Places" 。利用它的想法可以充分地挖掘地图查询日志的价值。其中的核心思想是当用户在地图上发起一个从地点A到地点B的查询时,表明用户对地点B感兴趣,尤其是当离A比较近时。
由于有大量的查询日志,可以从中发掘出很有用的信息。
论文的摘要部分:
Studies find that at least 20% of webqueries have local intent; and the fraction of queries with local intent thatoriginate from mobile properties may be twice as high. The emergence ofstandardized support for location providers in web browsers, as well asofproviders of accurate locations, enables socalled hyperlocal webqueryingwhere the location of a user is accurate at a much finergranularity than withIPbased positioning.
This paper addresses the problem ofdetermining the importanceof points of interest, or places, in localsearchresults. In doingso, the paper proposes techniques that exploit loggeddirectionsqueries. A query that asks for directions from a location a to a locationb is taken to suggest that a user is interested in traveling to band thus is avote that location b is interesting. Such usergenerated directions queries areparticularly interesting because they are numerous and contain preciselocations.
Specifically, the paper proposes aframework that takes a user location and a collection of nearby places asarguments, producinga ranking of the places. The framework enables a range ofaspectsof directions queries to be exploited for the ranking of places, includingthe frequency with which places have been referred to indirections queries.Next, the paper proposes an algorithm and accompanying data structures capableof ranking places in responseto hyperlocal web queries. Finally, an empiricalstudy with verylarge directions query logs offers insight into the potential ofdirections queries for the ranking of places and suggests that the proposedalgorithm is suitable for use in real web search engines.
具体的请看论文吧。