The Effect of Topic Pattern of Teen User's Search Behavior on Query Recommendation

Teenager users apply a limited vocabulary when they proceed to look for their desired materials. Another important issue is that teenagers often click mostly on the first items presented in the list of the search results. This research shows that, in order to amend and compensate these issues, we can extract and suggest a more appropriate query to the teenager user. This can be accomplished by discovering the relevant subject patterns from the behavior of the teenage user according to his or her previous search quarries and based on the already found patterns. In the proposed method, the topic patterns of the user are discovered based on the popularity of the clicks and the most relevant topics from the search logs which are generally massive. Afterwards, by using the binary classification method, the closest query to the query given by the user would be specified. Then, by filtering the subject navigation noise via extraction of the subject patterns of the teen user’s clicks, a user model with a higher accuracy can be obtained. We evaluated performance of the proposed method using the Alteryx and Weka tools, over the AOL search log, which includes about twenty million sample search transactions from six hundred and fifty different users. The results obtained from the experiments indicate that the queries presented by the proposed system are closer to the target user's query, and consequently, leads to achievement of more related results.