Semantic merge for feature requests: why similarity is not enough

What semantic similarity actually measures When a request is embedded as a vector, similarity is a number. It does not know whether two requests are duplicates. It knows whether their embeddings point