In the cultural heritage domain, there is a huge interest in utilizing semantic web technology and build services enabling users to query, explore and access the vast body of cultural heritage information that has been created over decades by memory institutions. For successful conversion of existing data into semantic web data, however, there is often a need to enhance and enrich the legacy data to validate and align it with other resources and reveal its full potential. In this visionary paper, we describe a framework for semantic enrichment that relies on the creation of thematic knowledge bases, i.e., about a given topic. These knowledge bases aggregate information by exploiting structured resources (e.g., Linked Open Data cloud) and by extracting new relationships from streams (e.g., Twitter) and textual documents (e.g., web pages). Our focused application in this paper is how this approach can be utilized when transforming library records into semantic web data based on the FRBR model in the process that commonly is called FRBRization.