Open chullhwan-song opened 5 years ago
standalone neural ranking model > retrieve documents from a large-scale collection
전통적인 term matching retrieval models > TF-IDF, BM25, query likelihood
여기서 말하는것은 이러한 전통적인 방법에 적용할 수 있는 feature representation 이란 이야기 그리고 sparse한 속성을 지니게.. >> SNRM
SNRM >> neural network model for learning standalone rankers based on latent sparse representations.
retrieval score = query(q) 와 document(d)
responsive standalone ranker 를 만들기 위해서 3가지 요소가 필요
input - > text query & text document
목표
network
training
inference
https://ciir-publications.cs.umass.edu/pub/web/getpdf.php?id=1302