In accordance with the xml-schemes of aggregating mathematical libraries, metadata sets must contain information about the scientific organization in which the study was conducted. Currently, this information is formalized as an affiliation of the authors. Affiliation is a mandatory attribute of a modern scientific publication. The affiliation may be incomplete and may contain outdated information. This does not allow to form the required set of metadata. In this paper, we propose methods for refining and replenishing the metadata of documents of electronic mathematical collections. We use the open Web-resources for refining and replenishing the authors' affiliation components. Using the software tools of the metadata factory of the Lobachevskii Digital Mathematical Library, the affiliation block is extracted, its components are parsed and analyzed. Semantic links of affiliation components with network information objects are created using the system of requests to open data sources developed by us. This made it possible to refine the metadata based on the subsequent analysis of the affiliation components. In the case when the necessary data is missing in the article, we replenish the components of the affiliation.