Artificial Intelligence (AI) is a key driver of the Industry 4.0 revolution. In industrial automation systems, data points of assets are represented by globally unique identifiers known as “Tags,” which often contain abbreviated asset and attribute information. These abbreviations need translation into concrete names to map data points to their corresponding assets. In this paper, we introduce DELA (Dual Embedding using LSTM and Attention), an innovative deep learning approach that uses two neural networks to classify “Attribute” and “Asset” for tag-to-asset mapping. The models are trained on real-world industrial standard datasets from the automation industry. To evaluate the generalization of our models, our experiments included a testing dataset with numerous abbreviations not present in the training set. This setup ensures that DELA can handle data with uncommon naming conventions. Our extensive experiments show that DELA efficiently achieves surpassing performance over current state-of-the-art approaches.