During the last three decades, information technologies are adopted by many libraries. It provides public access to their material in digital form to improve service. Metadata are the key aspect that must be considered to achieve a proper integration of digital library. It is data about data and has many purposes: Data description, data browsing and data transfer. The advanced search engine for text documents allowed retrieving text information in an efficient way. For the organization structured digital collections on internet scale, metadata is an approach for retrieval improvement, preservation, and interoperability. However, such engines experienced low accuracy when documents had unique properties that need specialized and deeper semantic extraction. By Combining the strengths of the deep learning models with that of word embedding is the key to high-performance metadata classification in digital library recommendation system. Throughout this article, we aim at providing a proposed method on the utilization of the deep machine learning approaches to build digital library recommendation system based on Metadata for Arabic and English languages.