Improving Search Rankings with Conversion Events

With semantic search, what are some best practices for incorporating human feedback (RHLF, eg. conversions) to improve the search rankings over time? How might rankings be improved with a vector database?

For example:

  1. User searches a keyword
  2. User clicks on an item
  3. Search engine moves most commonly clicked on items up in the rankings
1 Like

Iā€™m also interested in how this could be implemented.

Perhaps using the metadata feature to associate keywords and the vector with positive and negative outcomes?