Current domain specific or task-oriented human-computer dialogue systems are tightly bound with applications. The success of the strategy in one domain often does not ensure the success of the same strategy in other domains. To apply a strategy of one domain to another can be as difficult as exploring in a new domain. We present a spoken dialogue model based on extended Lambek calculus, which aims to solve the problem of understating complex sentences in a dialogue. This may pave the way towards applying a common strategy in different domains.