Generating in-app guided edits including concise instructions and coachmarks

1. A non-transitory computer readable medium comprising instructions that, when executed by at least one processor, cause a computing device to:

identify, from a repository of descriptions for guiding a user through operations associated with a client application, a description of an operation to perform within the client application;

generate, from encodings extracted from the description of the operation utilizing multi-task summarization neural network comprising a first set of layers for a concise instruction generator and a second set of layers for a coachmark generator:

a concise instruction by utilizing the first set of layers for the concise instruction generator of the multi-task summarization neural network to:

generate, a sequence of vector representations corresponding to words from the description of the operation; and

determine, based on the sequence of vector representations, copy probabilities and generation probabilities for successive time steps associated with the multi-task summarization neural network;

a coachmark by utilizing the second set of layers for the coachmark generator of the multi-task summarization neural network to:

determine, for a plurality of coachmark identifiers, respective probabilities of corresponding to the description of the operation; and

identify, from the plurality of coachmark identifiers; and

provide, for display within a user interface of the client application, representations of the concise instruction and the coachmark.

2. The non-transitory computer readable medium of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the computing device to generate the coachmark by utilizing the second set of layers for the coachmark generator of the multi-task summarization neural network generate the coachmark by identifying a user interface element selectable for performing the operation within the client application.

3. The non-transitory computer readable medium of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the computing device to generate the concise instruction by utilizing the first set of layers for the concise instruction generator of the multi-task summarization neural network to:

determine the copy probabilities by determining likelihoods of selecting words from the description of the operation to utilize within the concise instruction; and

determine the generation probabilities by determining likelihoods of selecting new words from a vocabulary to utilize within the concise instruction.

4. The non-transitory computer readable medium of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the computing device to select words for the concise instruction based on the copy probabilities and the generation probabilities.

5. The non-transitory computer readable medium of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the computing device to jointly generate the coachmark and the concise instruction utilizing the first set of layers and the second set of layers of the multi-task summarization neural network.

6. The non-transitory computer readable medium of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the computing device to jointly generate multiple coachmarks for multiple operations based on determining relationships between the multiple coachmarks utilizing the first set of layers and the second set of layers of the multi-task summarization neural network.

7. The non-transitory computer readable medium of claim 1 , wherein the multi-task summarization neural network comprises an encoder, a decoder, the second set of layers for the coachmark generator, and the first set of layers for the concise instruction generator.

8. The non-transitory computer readable medium of claim 7 , further comprising instructions that, when executed by the at least one processor, cause the computing device to generate the concise instruction by conditioning the decoder based on the coachmark.

9. A system comprising:

one or more memory devices comprising a repository of descriptions for guiding a user through operations associated with a client application, a plurality of coachmarks corresponding to respective user interface elements associated with the client application, and a multi-task summarization neural network comprising an encoder, a decoder, a first set of layers for a coachmark generator, and a second set of layers for a concise instruction generator; and

one or more computing devices that are configured to cause the system to:

identify, from the repository of descriptions, a description of an operation to perform within the client application;

determine, from the plurality of coachmarks, a coachmark that indicates a user interface element associated with the description of the operation by:

utilizing the encoder to generate encoder vector representations of the description; and

utilizing the first set of layers for the coachmark generator of the multi-task summarization neural network to determine, for the encoder vector representations, respective probabilities of corresponding to the description of the operation;

generate a concise instruction corresponding to the coachmark by:

utilizing the decoder to generate decoder vector representations based on the encoder vector representations; and

utilizing the second set of layers for the concise instruction generator of the multi-task summarization neural network to determine, based on the decoder vector representations, copy probabilities and generation probabilities for successive time steps associated with the multi-task summarization neural network; and

provide, for display within a user interface associated with the client application, representations of the concise instruction and the coachmark.

10. The system of claim 9 , wherein the one or more computing devices are further configured to cause the system to generate the concise instruction to include one or more words that correspond to the coachmark.

11. The system of claim 9 , wherein the one or more computing devices are configured to cause the system to generate the concise instruction to include a first word from the description of the operation and a second generated word from a vocabulary associated with the multi-task summarization neural network.

12. The system of claim 9 , wherein the one or more computing devices are configured to cause the system to utilize the concise instruction generator to generate the respective probabilities of corresponding to the description of the operation based further on the coachmark.

13. The system of claim 9 , wherein the one or more computing devices are configured to cause the system to determine the coachmark by determining that the coachmark has a highest probability from among the plurality of coachmarks.

14. The system of claim 9 , wherein the one or more computing devices are configured to further cause the system to utilize the multi-task summarization neural network to generate a second coachmark for a second operation independent from the coachmark associated with the operation.

15. The system of claim 9 , wherein the one or more computing devices are configured to further cause the system to utilize the multi-task summarization neural network to jointly generate an additional coachmark for an additional operation based on determining a relationship between the coachmark and the additional coachmark.

16. A computer-implemented method for providing in-app instructions for guiding users through tasks associated with client applications, the computer-implemented method comprising:

identifying, from a repository of descriptions for guiding a user through operations associated with a client application, a description of an operation to perform within the client application;

generating, from encodings extracted from the description of the operation utilizing a multi-task summarization neural network comprising a first set of layers for a concise instruction generator and a second set of layers for a coachmark generator:

a concise instruction by utilizing the first set of layers for the concise instruction generator of the multi-task summarization neural network to:

generate a sequence of vector representations corresponding to words from the description of the operation; and

determine, based on the sequence of vector representations, copy probabilities and generation probabilities for successive time steps associated with the multi-task summarization neural network; and

a coachmark by utilizing the second set of layers for the coachmark generator of the multi-task summarization neural network to:

determine, for a plurality of coachmark identifiers, respective probabilities of corresponding to the description of the operation; and

identify, from the plurality of coachmark identifiers, a coachmark identifier with a highest probability; and

providing, for display within a user interface associated with the client application, representations of the concise instruction and the coachmark.

17. The computer-implemented method of claim 16 , wherein the concise instruction comprises a summarized version of the description of the operation.

18. The computer-implemented method of claim 16 , wherein the coachmark comprises a visual indicator for display to indicate a particular user interface element that a user is to interact with as part of performing the operation.

19. The computer-implemented method of claim 16 , wherein providing the representations of the concise instruction and the coachmark comprises:

providing the coachmark for display in a position to outline a user interface element within the user interface; and

providing the concise instruction for display in a position within the user interface based on the position of the coachmark.

20. The computer-implemented method of claim 16 , further comprising:

receiving a request to access a tutorial for performing the operation from a client device running the client application; and

providing the representations of the concise instruction and the coachmark to the client device in response to the request.