Part 2 Learning domain-specific intent

In part 1, you learned the mechanics of matching and ranking on keywords (using TF-IDF and BM25) and numerical vectors (using cosine and dot product). You were also introduced to an overview of crowdsourced relevance ranking. Before performing these ranking techniques, however, it’s important to be able to correctly interpret a user’s query and run an appropriate search that understands the user’s intent. If an incorrect query is run and the wrong documents are matched, no amount of ranking logic on those bad results is going to overcome a misinterpreted query.