Granted Patent B2

US 12,412,022 · App. 17/542,844 · Granted Sep 9, 2025

Inventors: Jessica Lundin (Bellevue, WA); Sönke Rohde (San Francisco, CA); Owen Winne Schoppe (Orinda, CA); Michael Sollami (Cambridge, MA); David Woodward (Bozeman, MT); Brian Lonsdorf (San Francisco, CA); Alan Martin Ross (San Francisco, CA)

Assignee: Salesforce, Inc.

G06F40/103G06F40/247G06N5/04G06T11/60

Abstract

Claims (67)

1. A computer-implemented method comprising:

receiving an input text;

extracting keywords from the input text;

generating representative keywords from the keywords;

generating a graph representation of the representative keywords, wherein the graph representation comprises nodes representing the keywords and edges connecting the nodes and wherein at least three of the nodes are connected cyclically;

receiving images associated with the representative keywords;

selecting a visual-representation style based on the graph representation of the representative keywords, wherein the selected visual-representation style comprises at least two disjoint visual elements; and

arranging the images associated with the representative keywords according to the selected visual-representation style and the graph representation of the representative keywords.

2. The computer-implemented method of claim 1 , wherein the input text comprises text of a communications channel, text of a speech, or meeting minutes.

3. The computer-implemented method of claim 1 , wherein extracting keywords from the input text comprises using term frequency-inverse document frequency (TF-IDF), TopicRank, Yet Another Keyword Extractor (YAKE), KeyBert, or Rapid Automated Keyword Extraction (RAKE).

4. The computer-implemented method of claim 1 , wherein generating representative keywords from the keywords further comprises grouping keywords into keyword groups using multi-word synonym inference and selecting the representative keywords from the keyword groups.

5. The computer-implemented method of claim 4 , wherein the representative keywords represent ideas, and further comprising:

generating idea counts for the ideas using the multi-word synonym inference; and

adjusting a visual property of the images associated with the representative keywords based on the idea counts for the ideas represented by the representative keywords.

6. The computer-implemented method of claim 1 , wherein receiving images associated with the representative keywords further comprises applying a selected style to the images using a generative adversarial network.

7. The computer-implemented method of claim 1 , further comprising receiving additional input text;

extracting new keywords from the additional input text;

generating new representative keywords from the new keywords and the keywords;

modifying the graph representation of the representative keywords based on the new representative keywords;

receiving images associated with the new representative keywords;

modifying the selection of the visual-representation style based on the modifying of the graph representation of the keywords; and

modifying the arrangement of the images associated with the representative keywords according to the selected visual-representation style and the graph representation of the representative keywords based on the modifying of the visual-representation style and the modifying of the graph representation of the keywords, comprising at least one of adding an image representing a new representative keyword to the arrangement of images and removing an image representing a representative keyword from the arrangement of images.

8. The computer-implemented method of claim 1 , wherein the graph representation comprises nodes associated with the representative keywords and edges based on the relationships between the representative keywords in the input text.

9. A computer-implemented system comprising:

a processor that receives an input text,

extracts keywords from the input text,

generates representative keywords from the keywords,

generates a graph representation of the representative keywords, wherein the graph representation comprises nodes representing the keywords and edges connecting the nodes and wherein at least three of the nodes are connected cyclically,

receives images associated with the representative keywords,

selects a visual-representation style based on the graph representation of the representative keywords, wherein the selected visual-representation style comprises at least two disjoint visual elements, and

arranges the images associated with the representative keywords according to the selected visual-representation style and the graph representation of the representative keywords.

10. The computer-implemented system of claim 9 , wherein the input text comprises text of a communications channel, text of a speech, or meeting minutes.

11. The computer-implemented system of claim 9 , wherein the processor extracts keywords from the input text comprises using term frequency-inverse document frequency (TF-IDF), TopicRank, Yet Another Keyword Extractor (YAKE), KeyBert, or Rapid Automated Keyword Extraction (RAKE).

12. The computer-implemented system of claim 9 , wherein the processor generates representative keywords from the keywords further by grouping keywords into keyword groups using multi-word synonym inference and selecting the representative keywords from the keyword groups.

13. The computer-implemented system of claim 12 , wherein the representative keywords represent ideas, wherein the processor further:

generates idea counts for the ideas using the multi-word synonym inference, and

adjusts a visual property of the images associated with the representative keywords based on the idea counts for the ideas represented by the representative keywords.

14. The computer-implemented system of claim 9 , wherein the processer receives images associated with the representative keywords by applying a selected style to the images using a generative adversarial network.

15. The computer-implemented system of claim 9 , wherein the processor further receives additional input text;

extracts new keywords from the additional input text,

generates new representative keywords from the new keywords and the keywords,

modifies the graph representation of the representative keywords based on the new representative keywords,

receives images associated with the new representative keywords,

modifies the selection of the visual-representation style based on the modifying of the graph representation of the keywords, and

modifies the arrangement of the images associated with the representative keywords according to the selected visual-representation style and the graph representation of the representative keywords based on the modifying of the visual-representation style and the modifying of the graph representation of the keywords, comprising at least one of adding an image representing a new representative keyword to the arrangement of images and removing an image representing a representative keyword from the arrangement of images.

16. The computer-implemented system of claim 9 , wherein the graph representation comprises nodes associated with the representative keywords and edges based on the relationships between the representative keywords in the input text.

17. A system comprising: one or more computers and one or more non-transitory storage devices storing instructions which are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:

receiving an input text;

extracting keywords from the input text;

generating representative keywords from the keywords;

generating a graph representation of the representative keywords, wherein the graph representation comprises nodes representing the keywords and edges connecting the nodes and wherein at least three of the nodes are connected cyclically;

receiving images associated with the representative keywords;

selecting a visual-representation style based on the graph representation of the representative keywords, wherein the selected visual-representation style comprises at least two disjoint visual elements; and

arranging the images associated with the representative keywords according to the selected visual-representation style and the graph representation of the representative keywords.

18. The system of claim 17 , the instructions which are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising generating representative keywords from the keywords further comprise instructions which are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:

grouping keywords into keyword groups using multi-word synonym inference and selecting the representative keywords from the keyword groups.

19. The system of claim 17 , wherein the representative keywords represent ideas, and wherein the one or more computers and one or more non-transitory storage devices further store instructions which are operable, when executed by the one or more computers, to cause the one or more computers to perform the operation comprising:

generating idea counts for the ideas using the multi-word synonym inference; and

adjusting a visual property of the images associated with the representative keywords based on the idea counts for the ideas represented by the representative keywords.

20. The system of claim 17 , wherein the one or more computers and one or more non-transitory storage devices further store instructions which are operable, when executed by the one or more computers, to cause the one or more computers to further perform operations comprising:

receiving additional input text;

extracting new keywords from the additional input text;

generating new representative keywords from the new keywords and the keywords;

modifying the graph representation of the representative keywords based on the new representative keywords;

receiving images associated with the new representative keywords;

modifying the selection of the visual-representation style based on the modifying of the graph representation of the keywords; and

modifying the arrangement of the images associated with the representative keywords according to the selected visual-representation style and the graph representation of the representative keywords based on the modifying of the visual-representation style and the modifying of the graph representation of the keywords, comprising at least one of adding an image representing a new representative keyword to the arrangement of images and removing an image representing a representative keyword from the arrangement of images.

Assignments (2)

Continuity (1)

References Cited (20)