Research in information retrieval (IR) has largely been directed towards tasks requiring high precision. Recently, other IR applications which can be described as recall-oriented IR tasks have received increased attention in the IR research domain. Prominent among these IR applications are patent search and legal search, where users are typically ready to check hundreds or possibly thousands of documents in order to find any possible relevant document. The main concerns in this kind of application are very different from those in standard precision-oriented IR tasks, where users tend to be focused on finding an answer to their information need that can typically be addressed by one or two relevant documents. For precision-oriented tasks, mean average precision continues to be used as the primary evaluation metric for almost all IR applications. For recall-oriented IR applications the nature of the search task, including objectives, users, queries, and document collections, is different from that of standard precision-oriented search tasks. In this research study, two dimensions in IR are explored for the recall-oriented patent search task. The study includes IR system evaluation and multilingual IR for patent search. In each of these dimensions, current IR techniques are studied and novel techniques developed especially for this kind of recall-oriented IR application are proposed and investigated experimentally in the context of patent retrieval. The techniques developed in this thesis provide a significant contribution toward evaluating the effectiveness of recall-oriented IR in general and particularly patent search, and improving the efficiency of multilingual search for this kind of task.