{"id":"arxiv/2610.05058","url":"https://arxiv.org/abs/2610.05058","tags":["stat.ML","cs.LG","math.ST"],"score":0,"title":"A Statistical Inference Framework for PMI Estimation and SGNS Word Embeddings","domain":"stat.ML","source":"arxiv","authors":["Zhongqi Fan"],"reviews":0,"subject":"stat.ML","summary":"Pointwise Mutual Information (PMI) is a core measure of testing word association, and Skip-gram with Negative Sampling (SGNS) is essentially a method that implicitly factorizes a shifted PMI matrix. However, a systematic and well-rounded characterization of finite-sample uncertainty in PMI estimation remains absent and imperative to venture into. We provide a statistical framework for PMI estimation and its connection to SGNS. We prove consistency, asymptotic unbiasedness, and asymptotic normality of the empirical PMI estimator, derive its variance via the Delta method, and, applying…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-10-04","reactions":{}}