Add semantic search explainability

[davidmezzetti]

There are a couple of great libraries for model explainability, SHAP and Transformers Interpret. These libraries primarily focus on explaining classification, question-answering, summarization and translation model outputs. There isn't an easy way to explain the output of embeddings models.

This issue will add explain methods to embeddings instances. The implementation will create n permutations for each input text element to determine term importance for query results.