This paper is concerned with how to improve the overall performance of chat-oriented dialogue system. This research is motivated by the fact that majority of current chat engines are based on pattern matching. The knowledge base of this type of systems is predefined pattern-answer pairs such as the categories defined in Artificial Intelligent Markup Language (AIML). The inherent disadvantage of this kind of chat engines is that the interaction is carried out without any syntactic, semantic and contextual information. We propose a chat engine which is capable of dynamic knowledge acquisition and inference for a higher level of conversation intelligence. The dialogue engine leverages on natural language processing tasks such as syntactic and semantic parsing, named entity recognition, dialogue act detection, polarity analysis, etc., as well as dialogue history and heuristic rules for analysis and inference to achieve better understanding and intelligence.