Discrete choice modeling and assortment optimization receive significant attentions in the revenue management community. Yet, the major focus has been on substitutable products from a single category in which a product’s perceived utility depends only on the product itself. In this talk, I will present two recent research projects to broaden the scope. In the first project, we introduce a new choice model, the "contextual multinomial logit" model, in which the utility of a presented item to the customer depends on what other items are offered beside it in an assortment. We show empirically that incorporating the context effects may significantly enhance the prediction scores compared with several widely used discrete choice models. In the second project, we analyze assortment optimization problems under the multivariate multinomial logit models and develop approximation algorithms with theoretical guarantees.