Machine-learning-based digital survey creation and management

1. A method comprising:

receiving, from an administrator device and via a survey creation graphical user interface, user input to create an initial survey question;

inputting the initial survey question into a survey-creation-machine learner to:

apply natural-language processing to identify textual features of the initial survey question comprising an intent of the initial survey question;

select a representative survey question for the initial survey question based on the identified textual features comprising the intent of the initial survey question; and

determine, from a correlation database and based on the representative survey question, a first suggested survey question corresponding to a first survey category and a second suggested survey question corresponding to a second survey category;

receiving, from the survey-creation-machine learner, the first suggested survey question and the second suggested survey question; and

providing, for display within the survey creation graphical user interface of the administrator device, a first selectable option for the first suggested survey question and a second selectable option for the second suggested survey question.

2. The method of claim 1 , wherein utilizing the survey-creation-machine learner to identify the textual features of the initial survey question comprises:

extracting terms from the initial survey question; or

determining an intent for the initial survey question.

3. The method of claim 2 , wherein utilizing the survey-creation-machine learner to select the representative survey question for the initial survey question comprises:

identifying the extracted terms within the representative survey question from among candidate-survey questions;

identifying synonymous terms within the representative survey question corresponding to the extracted terms from the initial survey question; or

determining that a reciprocal intent of the representative survey question corresponds to the intent of the initial survey question.

4. The method of claim 1 , further comprising:

inputting training survey responses corresponding to a demographic group to a suggested-action-machine learner;

utilizing the suggested-action-machine learner to identify suggested-training-action items for the demographic group;

training the suggested-action-machine learner to select suggested-action items for the demographic group by comparing the suggested-training-action items to ground-truth-action items corresponding to the demographic group; and

utilizing the suggested-action-machine learner to determine a suggested-action item based on a survey response.

5. The method of claim 1 , further comprising:

receiving an indication from the administrator device of a user selection of the first selectable option for the first suggested survey question;

identifying supplementary suggested survey questions corresponding to the first survey category; and

providing the supplementary suggested survey questions for display within an updated survey creation graphical user interface of the administrator device.

6. The method of claim 1 , further comprising, before receiving the user input to create the initial survey question:

inputting a training survey question into the survey-creation-machine learner;

utilizing the survey-creation-machine learner to:

identify training textual features of the training survey question; and

select a candidate-representative-survey question for the training survey question based on the identified training textual features; and

training the survey-creation-machine learner to select representative survey questions for initial survey questions by comparing the candidate-representative-survey question to a ground-truth-representative-survey question.

7. The method of claim 6 , wherein training the survey-creation-machine learner to select the representative survey questions for the initial survey questions comprises:

determining a loss from a loss function based on comparing the candidate-representative-survey question to the ground-truth-representative-survey question; and

adjusting machine-learning parameters of the survey-creation-machine learner based on an objective to decrease the loss in a subsequent iteration.

8. The method of claim 1 , wherein the survey-creation-machine learner comprises a recurrent neural network or a Naïve Bayes Support Vector Machine.

9. A system comprising:

at least one processor; and

at least one non-transitory computer readable storage medium storing instructions that, when executed by the at least one processor, cause the system to:

receive, from an administrator device and via a survey creation graphical user interface, user input to create an initial survey question;

utilize a survey-creation-machine learner to:

apply natural-language processing to identify textual features of the initial survey question comprising an intent of the initial survey question;

select a representative survey question for the initial survey question based on the identified textual features comprising the intent of the initial survey question; and

determine, from a correlation database and based on the representative survey question, a first suggested survey question corresponding to a first survey category and a second suggested survey question corresponding to a second survey category;

receive, from the survey-creation-machine learner, the first suggested survey question and the second suggested survey question; and

provide, for display within the survey creation graphical user interface of the administrator device, a first selectable option for the first suggested survey question and a second selectable option for the second suggested survey question.

10. The system of claim 9 , further comprising instructions that, when executed by the at least one processor, cause the system to:

receive, from the administrator device, demographic indicators for target survey recipients, a recipient location for the target survey recipients, and a time range in which to send the target survey recipients the initial survey question and the first suggested survey question;

utilize a survey-timeframe-machine learner to determine from within the time range a suggested timeframe in which to send the initial survey question and the first suggested survey question to the target survey recipients, the suggested timeframe corresponding to a highest predicted response rate for the target survey recipients; and

provide the suggested timeframe for display within a graphical user interface of the administrator device.

11. The system of claim 10 , further comprising instructions that, when executed by the at least one processor, cause the system to utilize the survey-timeframe-machine learner to determine the suggested timeframe by:

determining a first response rate for a first survey cluster comprising responses from prior survey recipients who received digital surveys within a first timeframe and who correspond to the demographic indicators and the recipient location for the target survey recipients;

determining a second response rate for a second survey cluster comprising responses from prior survey recipients who received digital surveys within a second timeframe and who correspond to the demographic indicators and the recipient location for the target survey recipients;

applying a first machine-learning parameter to the first response rate to determine a first weighted response rate; and

applying a second machine-learning parameter to the second response rate to determine a second weighted response rate.

12. The system of claim 11 , further comprising instructions that, when executed by the at least one processor, cause the system to utilize the survey-timeframe-machine learner to determine the suggested timeframe by:

determining that the first weighted response rate exceeds the second weighted response rate; and

selecting a current corollary of the first timeframe as the suggested timeframe based on the first weighted response rate exceeding the second weighted response rate.

13. The system of claim 10 , further comprising instructions that, when executed by the at least one processor, cause the system to, before receiving the demographic indicators, the recipient location, and the time range:

input, into the survey-timeframe-machine learner, training demographic indicators for prior survey recipients, a training recipient location for the prior survey recipients, and a training time range in which the prior survey recipients received survey questions;

utilize the survey-timeframe-machine learner to determine from within the training time range a training timeframe for sending the survey questions; and

train the survey-timeframe-machine learner to determine suggested timeframes for sending suggested survey questions to survey recipients of identified demographic indicators within identified recipient locations by comparing the training timeframe to a ground-truth timeframe for sending the survey questions, the ground-truth timeframe corresponding to a highest recorded response rate for the prior survey recipients.

14. The system of claim 13 , further comprising instructions that, when executed by the at least one processor, cause the system to train the survey-timeframe-machine learner to determine the suggested timeframes for sending the suggested survey questions to the survey recipients by:

determining a loss from a loss function based on comparing the training timeframe to the ground-truth timeframe; and

adjusting machine-learning parameters of the survey-timeframe-machine learner based on an objective to decrease the loss in a subsequent iteration.

15. A non-transitory computer readable medium storing instructions thereon that, when executed by at least one processor, cause a computing system to:

receive, from an administrator device and via a survey creation graphical user interface, user input to create an initial survey question;

utilize a survey-creation-machine learner to:

apply natural-language processing to identify textual features of the initial survey question comprising an intent of the initial survey question;

select a representative survey question for the initial survey question based on the identified textual features comprising the intent of the initial survey question; and

determine, from a correlation database and based on the representative survey question, a first suggested survey question corresponding to a first survey category and a second suggested survey question corresponding to a second survey category;

receive, from the survey-creation-machine learner, the first suggested survey question and the second suggested survey question; and

provide for display within the survey creation graphical user interface of the administrator device, a first selectable option for the first suggested survey question and a second selectable option for the second suggested survey question.

16. The non-transitory computer readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the computing system to:

provide the first suggested survey question to recipient devices associated with survey recipients;

receive a response to the first suggested survey question from a recipient device of the recipient devices; and

utilize a suggested-action-machine learner to determine a suggested action item based on the response.

17. The non-transitory computer readable medium of claim 16 , further comprising instructions that, when executed by the at least one processor, cause the computing system to utilize the suggested-action-machine learner to determine the suggested action item by:

determining a suggested follow-up survey question;

determining a suggested follow-up contact;

determining a suggested meeting with a survey recipient associated with the recipient device; or

determining a suggested strategy, tactic, or initiative for a manager or team of the survey recipient associated with the recipient device.

18. The non-transitory computer readable medium of claim 16 , further comprising instructions that, when executed by the at least one processor, cause the computing system to utilize the suggested-action-machine learner to determine the suggested action item by:

applying natural-language processing to identify terms within the response; and

determining the suggested action item based on the response and the terms.

19. The non-transitory computer readable medium of claim 16 , further comprising instructions that, when executed by the at least one processor, cause the computing system to, before receipt of the user input to create the initial survey question:

input, into the suggested-action-machine learner, a training response to a survey question from a training survey respondent;

utilize the suggested-action-machine learner to determine a suggested-training-action item based on the training response; and

train the suggested-action-machine learner to determine suggested action items for responses by determining a loss from a loss function based on comparing the suggested-training-action item to a ground-truth-action item for the training response and adjusting machine-learning parameters of the suggested-action-machine learner based on an object to decrease the loss in a subsequent iteration.

20. The non-transitory computer readable medium of claim 16 , further comprising instructions that, when executed by the at least one processor, cause the computing system to:

receive a demographic indicator for a survey recipient who provided the response to the first suggested survey question and an action-item-completion indicator indicating that the survey recipient completed or failed to complete the suggested action item; and

utilize an additional suggested-action-machine learner to determine a suggested-follow-up-action item based on the demographic indicator and the action-item-completion indicator.