The paper examines a choice problem in case of large number of alternatives characterized by a set of criteria. Very often existing choice procedures cannot be used due to their computational complexity. In contrast, the approach proposed in this paper utilizes a set of easy to use techniques having complexity less than the quadratic. We apply super-threshold procedures sequentially in order to find the best alternatives. The application to learning to rank problem is also considered. Finally, two methods of defining thresholds and order of criteria in super-threshold procedures are studied. While the first method is based on the etalon criterion, the second method uses the distribution function. Both methods developed are mainly oriented on the choice of search results in search engines and capable of processing large amount of data in networks, but can also be applied to any system in which the alternatives are represented by a set of criteria.