US 12,462,169 B2
Machine learning techniques for generating visualization recommendations
Ryan Rossi, Santa Clara, CA (US); Xin Qian, Greenbelt, MD (US); Tak Yeon Lee, Cupertino, CA (US); Sungchul Kim, San Jose, CA (US); Sana Lee, Brea, CA (US); Fan Du, Milpitas, CA (US); and Eunyee Koh, San Jose, CA (US)
Assigned to ADOBE INC., San Jose, CA (US)
Filed by Adobe Inc., San Jose, CA (US)
Filed on Mar. 22, 2021, as Appl. No. 17/207,959.
Prior Publication US 2022/0300836 A1, Sep. 22, 2022
Int. Cl. G06N 20/00 (2019.01); G06F 18/213 (2023.01); G06N 5/04 (2023.01)
CPC G06N 5/04 (2013.01) [G06F 18/213 (2023.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A system for generating a recommended visualization of a dataset, the system comprising:
a processing device; and
a memory device in which instructions executable by the processing device are stored for configuring the processing device to perform operations including:
generating, from data attributes of an input dataset, a set of data attribute combinations;
modifying a meta-feature space that describes training meta-features of training datasets, wherein the modified meta-feature space includes first vector data describing i) meta-features extracted from the data attributes and ii) the training meta-features;
generating a set of multiple dense meta-feature vectors that includes, for each particular data attribute combination in the set of data attribute combinations, a respective dense meta-feature vector that includes respective meta-features describing the particular data attribute combination from the set of data attribute combinations;
generating a sparse meta-feature vector identifying a frequency of occurrences of the respective meta-features within the set of multiple dense meta-feature vectors;
identifying a set of visualization configurations, each visualization configuration including respective configuration attributes that describe visual characteristics of a dataset visualization;
accessing a configuration space that describes the set of visualization configurations, wherein the configuration space includes second vector data describing, for the each visualization configuration in the set of visualization configurations, the respective configuration attributes included in the each visualization configuration;
accessing a visualization scoring model that includes a wide scoring model and a deep scoring model, wherein the visualization scoring model is trained based on a visualization space that includes third vector data describing relationships of i) the training meta-features described by the meta-feature space with ii) the respective configuration attributes included in the each visualization configuration in the set of visualization configurations described by the configuration space;
generating (i) a dense configuration attribute set identifying the respective configuration attributes included in the each visualization configuration in the set of visualization configurations and (ii) a sparse configuration attribute set identifying a frequency of the respective configuration attributes in the dense configuration attribute set;
calculating, by the visualization scoring model and for the each visualization configuration in the set of visualization configurations, a respective recommendation score that is a combination of a respective wide score calculated by the wide scoring model and a respective deep score calculated by the deep scoring model;
identifying, based on the respective recommendation score for the each visualization configuration in the set of visualization configurations, a particular recommended visualization having a combination of (i) a particular visualization configuration from the set of visualization configurations and (ii) a particular meta-feature describing a particular data attribute combination from the dense meta-feature vector;
generating visualization image data based on the particular recommended visualization; and
causing a user interface to display the visualization image data.