Merchandise Recommendation for Retail Events with Word Embedding Weighted Tf-idf and Dynamic Query Expansion

We rank all we rely on item retrieval from marketplace inventory. With retrieved items by the sum of tf-idf scores from matched words, feedback to expand query scope, we discuss keyword expansion and keep the items with total tf-idf scores above a threshold. The candidate selection using word embedding similarity, and an retrieval based system works well to discover relevant enhanced tf-idf formula for expanded words in search ranking.