An improved noisy channel model to start with: https://aclanthology.org/D09-1093.pdf
a. Statistical Approach (including word2vec) b. Similar to semantic sequential labeling: also can use other tags: NER, POS, etc.
"red canada goose coat" (B-color,B-brand,I-brand,B-product)
good use for autocompletion and query rewritting
info about recent clicks/refinement
pseudo relevance feedback
simulates relevance feedback by reinforcing the top-ranked results as if the searcher had provided positive relevance feedback for them.
info about demographics - autocompletion
Most popular completion: lookup recent query history matching prefix input Generative Models: prob n-gram/LM w/o log