DeepEye通过关键字搜索可视化数据.pdf

DEEPEYE : Visualizing Your Data by Keyword Search Xuedi Qin† , Yuyu Luo† , Nan Tang ‡ , Guoliang Li† † ‡ Tsinghua University , Qatar Computing Research Institute Outline Background Key Intuitions The Architecture of DeepEye Visualization Search Engine (VSE) DeepEye Demo Background: Automatic Data Visualization Let usfirst review the automatic visualization recommendation process. 1.Data Transformation Background: Automatic Data Visualization 2. Enumerate All Valid Visualizations …… Background: Automatic Data Visualization 3. Rank Visualizations Why ? How about users who want to know the comparison of electricity usage in different cities? Visualization Recommendation systems may mislead the user, by generating visualizations that might be worse than nothing, since it is basically impossible to guess a user’s query intent from nothing. The natural problem arises: What is the most feasible way to (automatically) create good visualizations? Outline Background Key Intuitions The Architecture of DeepEye Visualization Search Engine (VSE) DeepEye Demo Key Intuitions Three key intuitions for handling the above problem (1) What is the ideal language for the user to specify his intent? Of course, the mother tongue – Google-like natural language search interface is friendly to everyone. Feasibility Analysis ： ①most visualizations are generated using declarative languages like Vega-Lite[1] ②using machine learning to convert text to SQL queries for relational databases has been prov