Deep Learning and Machine Learning based models have become extremely popular in text processing and information retrieval, owing to their remarkable effectiveness. However, the complex non-linear structures underlying these models make them largely inscrutable. A significant body of research has focused on increasing the transparency of these models. This article provides a broad overview of research on the explainability and interpretability of information retrieval methods, focusing primarily on both neural models for document ranking, and retrieval-augmented generation systems. A significant part of this research is inspired by, and strongly related to, similar work done in the area of natural language processing (NLP). For completeness, therefore, we also briefly review (in the Appendices) some recent studies from the NLP community on explaining word embeddings, sequence modeling, attention modules, transformers, and BERT. The concluding section suggests some possible directions for future research on this topic.
Saha et al. (Thu,) studied this question.