Analyses of the spoken material of old sound newsreels offer great potential for insight into the way people spoke many decades ago. However, from a researcher’s point of view, newsreels represent a highly diverse and unstructured body of speech recordings, making it difficult to identify the best recordings for individual studies. Previous research highlighted the importance of using databases for this purpose. A crucial aspect of database development is the establishment of a metadata structure, which, providing keywords for searching recordings, facilitates research by the identification of the recordings for a given research project. The present study examines the possibility of using some of the factors described by Dell Hymes in his model of speech communication for metadata structure development. Qualitative analysis was performed on the spoken material of the news stories produced in 1931 and 1932 in Hungary, with a focus on speech situations, speech events and participants. The research revealed a variety of categories belonging to these factors. It is concluded that the factors Hymes described may provide a good basis for metadata structure development; however, these factors seem to be too general, and addition of subcategories is needed for a more accurate description of speech events.