This document discusses probabilistic models for information retrieval. It motivates using a probabilistic framework to model relevance as depending on query and document representations. The goal is to find an ideal answer set R that maximizes the probability of relevance to the user. Documents in R are predicted relevant, others non-relevant. Bayes' rule is used to define the probabilistic model, making independence assumptions. Initial probabilities are adjusted to address small values. The probabilistic model ranks documents by decreasing probability of relevance. A brief comparison is made of classic Boolean and vector space models.
????? ??? ??? ???? ??? ?? ?? ??.
?? ?? ??.
- Social Network Social Media ??
- Social Media ???? ??
- Social Network Trend? ??? ??
- Social Media PR & Marketing ???
- Social Media ??
- ??? ?? Social Media
- Social Media ?? Point
- eStoryLAB ????