An empirical and analytic comparison of term-independence retrieval models for ad hoc retrieval

What is it about?

A more upto date comparison of (17) term-independence retrieval models including SPUD language model, MATF-like retrieval models, machine generated retrieval models, divergence from independence retrieval models and the common models (like BM25, language model and PL2). The best performing model is used as a baseline to compare based on a one-tail test. No model consistently performing without statistical significance difference from the top performing model across all the test collections using a multiple-hypothesis testing method.

Featured Image

Photo by averie woodard on Unsplash

Photo by averie woodard on Unsplash

Why is it important?

Some models do perform well in most test collections and perhaps those models should be used as the baseline for future comparison.

Perspectives