EXAMINING PERSONALIZATION HEURISTICS BY TOPICAL ANALYSIS OF QUERY LOG
EXAMINING PERSONALIZATION HEURISTICS BY TOPICAL ANALYSIS OF QUERY LOG
复制标题
DOI：
--
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
Wei Song-;Ying Liu;Li-zhen Liu;Hanshi Wang
中科院分区：
文献类型：
--
作者：
Wei Song-;Ying Liu;Li-zhen Liu;Hanshi Wang
. Personalization has been considered to be a promising way for future web search. Many algorithms have been proposed based on different heuristics. There is little work on systematically examining such heuristics. In this article, we examine typical personalization heuristics based on a commercial search engine query log. We aim to provide evidence to show the necessity, potential and limitations for personalization. Specially, we utilize automated query topic classification to simulate query intents and user preferences, which ensures that our study can be conducted on large scale query log data. We examine the heuristics on query ambiguity, query intent, user preferences and their interactions. We find that only a few queries with multiple user intents are inherently ambiguous. Queries with a specific meaning also have various user intents on different query aspects. About 48% of such queries involve named entities. We introduce new metrics to measure user preferences and preference distribution. These metrics show that searchers do have their preferences and focus on only a few interested topics. We assess the potential and difficulty for personalization based on search history. The results show that users with long search history benefit more from personalization, while the difficulty is nearly the same for users with either long or short search history. These observations could provide references for potential directions of web personalization.
