Retriever: An Agent for Intelligent Information Recovery

Abstract

Since the Internet and the amount of information published on it have grown at an exponential rate, finding relevant information has become a laborious and time-consuming process. While search engines were created to assist individuals in managing 
this overwhelming amount of information, they do have some drawbacks. Here we detail our current efforts towards developing Retriever, a self-sufficient agent that can process user requests and provide them with accurate responses. In order to begin its 
independent investigation of the Web, Retriever makes use of preexisting search engines to get beginning points. Afterwards, it undergoes a self-training procedure to gain knowledge of the query domain and enhance its performance. Upon learning the 
query domain, the agent refines its search approach, broadens its initial inquiry, and sets out to find the user-requested content.