Googler Pandu Nayak's interview notes which DOJ published on 2th May
justice.gov/atr/us-and-plain… have some really interesting points.
- Q* (pronounced "Q star"), Google's measure of quality of a document.
- Google have an internal debugger window which they can use to see signal weights for each document for a particular query.
- Google's traditional approach to ranking was in the style of Okapi BM25, a ranking function used to estimate the relevance of documents to a given search query.
- RankEmbed is a dual encoder model that embeds both query and document into embedding space. Embedding space considers semantic properties of query and document in addition to other signals. Retrieval and ranking are then a dot product (distance measure in the embedding space).
- Navboost is described as a QD table, a query-to-document lookup table, used in both directions containing counts/frequencies of user query activity by document.