Knowledge-based approach has drawn more and more attention in question answering since 2002. This approach incorporates syntax-based semantic information, sophisticated lexical and common sense reasoning, which can help to extract the exact answer based on the better understanding of the texts rather than some statistics and rules. In this presentation, firstly, I will survey the state of the art of the knowledge-based approach in QA. Secondly, I will propose my answer extraction method based on the syntactic/semantic representation. In my work, the semantic information is represented by predicate argument structures which will be generated automatically. The possible strength of this method is that we can extract answers for various types/categories of questions using a uniform representation and it can be easily adapted to a new domain/question type. Finally, some primary experimental results will be reported.