Online reviews are valuable sources of information to help companies to make good decisions for business intelligence. In this study, we propose an Automatic Aspect-based Sentiment Summarization (AAbSS) system that has two components and can generate a summary as an output. The first component is the Aspect-based Knowledge Representation and Selection (AKRS) used to represent reviews based on aspects and their polarities for selecting aspect-based knowledge. To represent and selection knowledge, a set of frequency of polarity opinion strength, a summation of frequency of aspect, and an information of aspect are initiated. The second component is the Summary Format Generation (SFG) used to automatically generate three kinds of formats. In this component (SFG), new representations for visual and structured summaries, and a new way of applied natural language generation for a textual summary are proposed. In the experiments, 13 domains from benchmark datasets of customer reviews, e.g. cell phone, digital camera, etc. are used. The proposed system not only fast generates summaries having good performance when compared to other summaries generated by other systems and easily updated when adding new reviews in the same domain but also does not spend memory capacity to save any raw data.