一、search engine and recommendation engine
1. search
key challenge: user-item senmantic gap
matching problem in search and recommendation (hector et al. CACM'11)
query-document matching
1.bridging the senmantic gap between words
2. the order of terms: 80%: choice of words, 20% order of words
relation between matching and ranking
in traditional IR: ranking =matching
in web search: matching is a feature for ranking
latent model: two latent spaces
translation model: one latent space
part 2: recommendation
traditional rec model: see the vedio
deep matching:
1.methods of representation learning:calculate the matching
2. method of matching function learning