Nowadays, Social communities are growing rapidly where opinions are expressed in natural language independently. Current question answering forums are a representation of future web search services like searching and posting queries or answers. Finding the significant answer to a recently posted query is expected by a framework, it furnishes the pool of answers with similar questions links, which could be an extended task. A solution to this problem is the framework provides a way to effectively rank the most relevant and best answers which are from historical archives based on similar queries found. It constructs training samples with positive, negative and neutral classes and then online component retrieves similar queries with their answer pools. Two approaches were compared to retrieve similar questions. The objective is to rank answer candidates based on pairwise comparison where question-answer pairs are ranked using an SVM-based rank approach based on an offline trained model which provides the user with most relevant answers for a given posted question.