Abstract: When facing multi-dimensions data,the challenge for users is to select appropriate attributes and create insightful visualizations.In order to explore more data attributes quickly and generate reasonable visualizations recommendations for different analysis tasks,we propose a new framework.Firstly,a set of visualization grammar rules is defined,and the rules of data transformation,aggregation and visual encoding are described.Also,we summarize the data analysis tasks and reference visualization analysis tasks,and construct relevant visualization recommendations.To help users explore data more flexibly and get insights in analysis,we provide an inter-active visualization system.Based on the visualization language and analysis task rules,the system generates visualization recommendations sets effectively according to users specified attributes and different analysis tasks.