Understanding User's Search Behavior towards Spiky Events

Abstract

Web searches are done by users every day on a million-daily basis. Many of these web searches are related to events, social occasions that attracts society's attention. Events may happen multiple times on cyclic or non-periodic occasions. These are known as spiky events. When these events occur, multiple spikes can be observed in query logs triggered by a change in the user's behaviour and an increase in the frequency of the user's queries. In this paper, we aim to understand the user's search behaviour towards this kind of events. To this regard, we propose a new taxonomy of spiky events which categorizes queries into two groups: periodic (ongoing, historical, traditional) and aperiodic (predictable and unpredictable), and study how various features concerning the query and the clicked web pages describe the user's behaviour, before, during, and after the event. To conduct this research, we consider 100 spiky events and rely on a two-year Persian search engine query log to analyse their related queries and associated information. The results obtained show that users have a different behaviour regarding the query frequency, length and temporality, depending on the category of the spiky event and that query formulation and clicked pages are also different for each category before, during and after the event. Understanding these user's behaviours and their relationship with the different categories may play an important role for any search engine looking to provide better services for their users.