qdrant-relevance-feedback

Improve search relevance by incorporating user feedback into the vector search process. Training a feedback model modifies retrieval behavior to enhance document relevance when dense retrievers underperform.

A relevance feedback model is a trained mechanism that modifies the vector search process based on user feedback. It enhances document retrieval quality by properly calibrating weights during search queries.

Train the feedback model using the relevance feedback API to modify the vector search process. This requires proper calibration of weights to enhance document retrieval quality and improve search recall.

Use a relevance feedback model for document retrieval when a dense retriever is underperforming. It addresses the challenge of improving search relevance and recall by incorporating user feedback into search queries.

Qdrant supports enhancing search quality through a Relevance Feedback API. This allows you to train a feedback model and apply it to search queries to modify the vector search process for improved relevance.

Limitations include the requirement to train a feedback model and properly calibrate weights. Applying this approach demands appropriate user feedback data to successfully modify the vector search process and improve retrieval.