The community member is asking if there is a way to customize the embedding of a node in the llama-index library, specifically by using only certain keywords ("how", "use", "llama-index") instead of the full node text. The comments suggest that while there may not be a direct option for customizing node text embeddings, the community members discuss using techniques like tokenization, stopword filtering, and query transformations to potentially improve the vector search results. They mention specific examples and notebooks from the llama-index project that may be helpful in exploring these approaches.
Is there a way to custom the embedding of a node? for example the node text is "how to use llama-index?" ,I want to just use "how,use,llama-index" to calculate the embedding?
