|
What is the best approach to predict user faceting by user query. I have search site with rich filtering/faciting model and based on user history beheviour I want automatically apply filters as a boost factors or hard filters. I wandering what is the best way to mathematecaly model such think. It is easy to calculate collocation(mutual info) between individual words and filters and final model can sum probabilities generated by each word, I think it is naive bayes like model... I want to generte prediction model from Words To Filters, perhaps machine translation would be helpful. Any ideas? |
