Each document field represents a different perspective on it's context, potentially highlighting a key area for the search system to focus on. You can embed each of these fields then calculate a weighted sum of the embeddings, also known as a Composite Embedding. A type of [[Document Embeddings]]. The composite embedding will allow the system to become more context aware, in addition to introducing another tunable hyper-parameter from controlling the search behavior. You can also define weights to the components of the composite embedding. The weightings allow you to assign priorities to each component based on your use case and the quality of your data. Intuitively, the size of these weightings is dependent on the semantic value of each component. Since the chunk text itself is by far the richest, you can start by assigning a weighting of 70%. The precise setting for these values has to be determined empirically, on a use-case by use-case basis. You can delete all the component embeddings to save space. [More details](https://www.elastic.co/search-labs/blog/advanced-rag-techniques-part-1#metadata-inclusion-and-generation)