Improving the Accuracy of Extracting Useful Information in Search Engines from the Web Using Deep Reinforcement Learning based on the Q-Learning Algorithm

Abstract

Due to the ever-increasing range of internet users as well as the exponential increase of various pages and contents to the web space, the amount of information available in the internet space has greatly increased and is also increasing. Therefore, increasing accuracy and speed in searching for information in search engines has become one of the most important research and practical aspects in the field of web information exploration. In this study, with the aim of modeling using the Q-learning algorithm, to increase the accuracy and speed of searching and to optimize the accuracy of extracting useful information. In the proposed method, data sets related to the characteristics of web pages are received as input.  The desired data is classified and labeled according to the investigated results, and then the similarity between the result and the text is done. This review was done with the Jaccard criterion and by considering a value as a threshold, the degree of similarity of the text and the result are categorized into two valuing classes 0 and 1. Finally, the numbered data along with the corresponding label are transferred to the Q-Learning algorithm.