Constructing content based on multi-sentence compression of source content

Patent metadata

Field	Value
Publication number	US-10949452-B2
Application number	US-201715854320-A
Country	US
Kind code	B2
Filing date	Dec 26, 2017
Priority date	Dec 26, 2017
Publication date	Mar 16, 2021
Grant date	Mar 16, 2021

Abstract

Embodiments of the present invention provide systems, methods, and computer storage media directed to facilitating corpus-based content generation, in particular, using graph-based multi-sentence compression to generate a final content output. In one embodiment, pre-existing source content is identified and retrieved from a corpus. The source content is then parsed into sentence tokens, mapped and weighted. The sentence tokens are further parsed into word tokens and weighted. The mapped word tokens are then compressed into candidate sentences to be used in a final content. The final content is assembled using ranked candidate sentences, such that the final content is organized to reduce information redundancy and optimize content cohesion.

First claim

What is claimed is:

1. One or more computer storage media storing computer-useable instructions that, when executed by one or more computing devices, causes the one or more computing devices to perform operations, the operations comprising:
obtaining source content relevant to an input snippet received via a user interface;
utilizing a graphical representation of a plurality of sentences from the source content to identify a set of sentences in the relevant source content having overlapping information, wherein the overlapping information is determined based on a set of keywords that are relevant to the input snippet received via the user interface;
generating a candidate sentence by compressing content in the set of sentences having overlapping information;
generating a final content comprising a set of candidate sentences including the candidate sentence; and
providing the final content as content automatically created in response to the input snippet.

2. The one or more computer storage media of claim 1 , further comprising parsing the source content relevant to the input snippet into a plurality of sentences.

3. The one or more computer storage media of claim 2 , wherein each of the plurality of sentences are mapped to a first graph and assigned a weight, wherein the weight relates to the relevance between the corresponding sentence and the input snippet received via the user interface.

4. The one or more computer storage media of claim 3 , wherein at least a portion of the weighted sentences are parsed into a plurality of word tokens.

5. The one or more computer storage media of claim 4 , wherein each of the plurality of word tokens are mapped to a second graph and assigned a weight, wherein the weight relates to the relevance between the corresponding word token and the input snippet received via the user interface.

6. The one or more computer storage media of claim 5 , wherein the weighted word tokens are compressed to generate the candidate sentence.

7. The one or more computer storage media of claim 1 , wherein the generation of the final content comprises sequencing at least a portion of the set of candidate sentences based on candidate sentence ranks, the final content assembled to reduce information coverage redundancy and optimize for overall coherence.

8. A computer implemented method for generating content based on graph-based sentence compression using retrieved source content present in a corpus, the method comprising:
obtaining source content relevant to an input snippet received via a user interface;
assigning a weight to each of a plurality of sentences from the source content based on a relevance between each of the plurality of sentences and the input snippet received via the user interface;
mapping each of the plurality of sentences to a first graph based on the respective weight of each of the plurality of sentences from the source content;
utilizing the first graph of the plurality of sentences from the source content to identify a set of sentences in the relevant source content having overlapping information, wherein the overlapping information is determined based on a set of keywords that are relevant to the input snippet;
generating a candidate sentence by compressing content in the set of sentences having overlapping information;
generating a final content comprising a set of candidate sentences including the candidate sentence; and
providing the final content as content automatically created in response to the input snippet.

9. The method of claim 8 , further comprising parsing the source content relevant to the input snippet into a plurality of sentences.

10. The method of claim 8 , wherein at least a portion of the weighted sentences are parsed into a plurality of word tokens, wherein each of the plurality of word tokens are mapped to a second graph and assigned a weight, wherein the weight relates to the relevance between the corresponding word token and the input snippet.

11. The method of claim 10 , wherein the weighted word tokens are compressed to generate the candidate sentence.

12. The method of claim 8 , wherein the generation of the final content comprises sequencing at least a portion of the set of candidate sentences based on candidate sentence ranks, the final content assembled to reduce information coverage redundancy and optimize for overall coherence.

13. The method of claim 8 , further comprising:
assigning a weight to the candidate sentence, the weight relating to the relevance between the candidate sentence and the input snippet;
reassigning an updated weight to each of the plurality of sentences based on the weight of the candidate sentence; and
mapping each of the plurality of sentences to a second graph based on the respective weight of each of the plurality of sentences.

14. The method of claim 8 , wherein the input snippet is received from a user via the user interface of an application of a user device.

15. A computer system comprising:
one or more processors; and
one or more non-transitory computer-readable storage media, coupled with the one or more processors, having instructions stored thereon, which, when executed by the one or more processors, cause the computing system to provide:
means for identifying a set of sentences in source content having overlapping information, the set of sentences being relevant to an input snippet received via a user interface, wherein the overlapping information is determined based on a set of keywords that are relevant to the input snippet received via the user interface;
means for generating a candidate sentence by compressing content in the set of sentences having overlapping information; and
means for generating a final content comprising a set of candidate sentences including the candidate sentence.

16. The system of claim 15 , further comprising means for parsing the source content relevant to the input snippet into a plurality of sentences.

17. The system of claim 16 , wherein each of the plurality of sentences are mapped to a first graph and assigned a weight, the weight relating to the relevance between a corresponding sentence and the input snippet.

18. The system of claim 17 , wherein the weighted sentences are parsed into a plurality of word tokens.

19. The system of claim 18 , wherein each of the word tokens are mapped to a second graph and assigned a weight, the weight relating to the relevance between the corresponding word token and the input snippet.

20. The system of claim 19 , wherein the weighted word tokens are compressed to generate the candidate content.

Assignees

• Adobe Inc

Inventors

• Srinivasan Balaji Vasan
• Maneriker Pranav Ravindra
• Modani Natwar
• Krishna Kundan

Classifications

• G06F40/205

  Parsing · CPC title

• G06F16/334 Primary

  Query execution (filtering based on additional data G06F16/335) · CPC title

• G06F40/284

  Lexical analysis, e.g. tokenisation or collocates · CPC title

• G06F16/338

  Presentation of query results · CPC title

• G06F40/151

  Transformation · CPC title