US 12,265,557 B2
Learning-based method for generating visualizations via natural language
William Brandon George, Pleasant Grove, UT (US); Wei Zhang, Great Falls, VA (US); Tyler Rasmussen, San Jose, CA (US); Tung Mai, San Jose, CA (US); Tong Yu, Fremont, CA (US); Sungchul Kim, San Jose, CA (US); Shunan Guo, San Jose, CA (US); Samuel Nephi Grigg, Sandy, UT (US); Said Kobeissi, Lovettsville, VA (US); Ryan Rossi, San Jose, CA (US); Ritwik Sinha, Cupertino, CA (US); Eunyee Koh, San Jose, CA (US); Prithvi Bhutani, Lynnwood, WA (US); Jordan Henson Walker, Lehi, UT (US); and Abhisek Trivedi, South Jordan, UT (US)
Assigned to Adobe Inc., San Jose, CA (US)
Filed by Adobe Inc., San Jose, CA (US)
Filed on Aug. 31, 2023, as Appl. No. 18/459,081.
Prior Publication US 2025/0077549 A1, Mar. 6, 2025
Int. Cl. G06F 40/00 (2020.01); G06F 16/242 (2019.01); G06F 16/28 (2019.01); G06F 40/205 (2020.01); G06F 40/40 (2020.01)
CPC G06F 16/287 (2019.01) [G06F 16/243 (2019.01); G06F 40/205 (2020.01); G06F 40/40 (2020.01)] 15 Claims
OG exemplary drawing
 
6. A computerized method performed by one or more processors, the method comprising:
accessing, by a natural language model training component, a first set of training data comprising a document corpus having text-based information corresponding to graphic visualizations, the text-based information including data attributes of the graphic visualizations;
training, by the natural language model training component, a natural language model on the first set of training data to determine word embeddings from n-gram inputs determined from a natural language request for generating a graphic visualization;
accessing, by a classifier training component, a second set of training data comprising labeled intent pairs, the labeled intent pairs including a text phrase associated with an intent label identifying a known intent of the text phrase;
training, by the classifier training component, an intent classifier on the second set of training data to determine a request intent responsive to an input comprising the natural language request; and
providing a graphic visualization generation component, the graphic visualization generator configured to:
receive a graphic visualization type selected based on a number of data-attribute embeddings mapped to the word embeddings output from the trained natural language model and from the request intent output from the trained intent classifier; and
generate a graphic visualization having a data attribute corresponding to the data-attribute embedding and in accordance with the graphic visualization type, the graphic visualization generated from a standard format structure using the having the graphic visualization type and the data attribute as structured data elements.