Emergency workflow trigger
Detection of crisis situations using multiple classifiers can enable fast and efficient processing of input phrases to determine if an abatement protocol should be executed. An input phrase can be passed through a pattern matching classifier and then through a trained machine learning classifier. If neither classifier identifies a crisis, the workflow can continue as usual. However, if a crisis is identified, confirmation of the crisis situation can be sought and used to further update one or both of the classifiers. If the crisis is confirmed, emergency information and crisis management tools can be presented to the user, among other mitigating actions. If the crisis is not confirmed, a prompt can be presented to the user to discuss the trigger phrase associated with the trigger signal.
1. A method, comprising:
obtaining training data comprising training input phrases, each training input phrase being associated with a label indicative of whether the respective training input phrase is associated with a crisis situation;
training a machine learning classifier using the training input phrases and their respective labels;
receiving an input phrase;
applying the input phrase to a machine learning classifier to determine whether the input phrase triggers the machine learning classifier;
determining a crisis score using the input phrase;
accessing one or more historical crisis scores; and
executing an abatement protocol in response to determining that the input phrase triggers the trained machine learning classifier, wherein execution of the abatement protocol is further based on the crisis score and the one or more historical crisis scores.
2. The method of claim 1 , wherein the machine learning classifier is a transformer-based machine learning classifier initially trained using pre-training training data for a language prior to the training the machine learning classifier using the training input phrases.
3. The method of claim 1 , further comprising:
determining a crisis score trend using the crisis score and the one or more historical crisis scores; wherein execution of the abatement protocol is further based on the crisis score trend.
4. The method of claim 1 , further comprising:
obtaining additional training data, the additional training data including one or more additional input phrases previously supplied to the trained machine learning classifier, each of the one or more additional input phrases being associated with a respective label indicative of whether a user confirmed or denied a respective crisis situation after the respective additional input phrase triggered the trained machine learning classifier, and
further training the trained machine learning classifier based at least in part on the additional training data, wherein further training occurs prior to receiving the input phrase.
5. The method of claim 4 , wherein each of the additional input phrases was previously supplied to the trained machine learning classifier and triggered the trained machine learning classifier.
6. The method of claim 1 , wherein executing the abatement protocol includes i) accessing and presenting a text string containing emergency contact information, ii) accessing and presenting a link configured to begin initiation of an emergency contact connection upon actuation; iii) sending a signal to automatically begin initiation of an emergency contact connection; iv) initiating a patient health questionnaire protocol; v) generating a prompt for selecting a crisis management tool; vi) automatically initiating a crisis management tool; or vii) any combination of i-vi.
7. The method of claim 1 , wherein receiving the input phrase occurs via a text-based communication protocol, wherein executing the abatement protocol includes automatically initiating a crisis management tool via the text-based communication protocol, the crisis management tool including a workflow of prompts and/or comments for mitigating the crisis situation.
8. The method of claim 7 , wherein the crisis management tool, when initiated, facilitates:
providing, via the text-based communication protocol, a series of questions from a patient health questionnaire; and
receiving, via the text-based communication protocol, a response for each question of the series of questions.
9. A system, comprising:
one or more data processors; and
a non-transitory computer-readable storage medium containing instructions which, when executed on the one or more data processors, cause the one or more data processors to perform operations including:
obtaining training data comprising training input phrases, each training input phrase being associated with a label indicative of whether the respective training input phrase is associated with a crisis situation;
training a machine learning classifier using the training input phrases and their respective labels;
receiving an input phrase;
applying the input phrase to a machine learning classifier to determine whether the input phrase triggers the machine learning classifier;
determining a crisis score using the input phrase;
accessing one or more historical crisis scores; and
executing an abatement protocol in response to determining that the input phrase triggers the trained machine learning classifier, wherein execution of the abatement protocol is further based on the crisis score and the one or more historical crisis scores.
10. The system of claim 9 , wherein the machine learning classifier is a transformer-based machine learning classifier initially trained using pre-training training data for a language prior to the training the machine learning classifier using the training input phrases.
11. The system of claim 9 , wherein the operations further include:
determining a crisis score trend using the crisis score and the one or more historical crisis scores; wherein execution of the abatement protocol is further based on the crisis score trend.
12. The system of claim 9 , wherein the operations further include:
obtaining additional training data, the additional training data including one or more additional input phrases previously supplied to the trained machine learning classifier, each of the one or more additional input phrases being associated with a respective label indicative of whether a user confirmed or denied a respective crisis situation after the respective additional input phrase triggered the trained machine learning classifier, and
further training the trained machine learning classifier based at least in part on the additional training data, wherein further training occurs prior to receiving the input phrase.
13. The system of claim 12 , wherein each of the additional input phrases was previously supplied to the trained machine learning classifier and triggered the trained machine learning classifier.
14. The system of claim 9 , wherein executing the abatement protocol includes i) accessing and presenting a text string containing emergency contact information, ii) accessing and presenting a link configured to begin initiation of an emergency contact connection upon actuation; iii) sending a signal to automatically begin initiation of an emergency contact connection; iv) initiating a patient health questionnaire protocol; v) generating a prompt for selecting a crisis management tool; vi) automatically initiating a crisis management tool; or vii) any combination of i-vi.
15. The system of claim 9 , wherein receiving the input phrase occurs via a text-based communication protocol, wherein executing the abatement protocol includes automatically initiating a crisis management tool via the text-based communication protocol, the crisis management tool including a workflow of prompts and/or comments for mitigating the crisis situation.
16. The system of claim 15 , wherein the crisis management tool, when initiated, facilitates:
providing, via the text-based communication protocol, a series of questions from a patient health questionnaire; and
receiving, via the text-based communication protocol, a response for each question of the series of questions.
17. A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause a data processing apparatus to perform operations including:
obtaining training data comprising training input phrases, each training input phrase being associated with a label indicative of whether the respective training input phrase is associated with a crisis situation;
training a machine learning classifier using of the training input phrases and their respective labels;
receiving an input phrase;
applying the input phrase to a machine learning classifier to determine whether the input phrase triggers the machine learning classifier;
determining a crisis score using the input phrase;
accessing one or more historical crisis scores; and
executing an abatement protocol in response to determining that the input phrase triggers the trained machine learning classifier, wherein execution of the abatement protocol is further based on the crisis score and the one or more historical crisis scores.
18. The computer-program product of claim 17 , wherein the machine learning classifier is a transformer-based machine learning classifier initially trained using pre-training training data for a language prior to the training the machine learning classifier using the training input phrases.
19. The computer-program product of claim 17 , wherein the operations further include:
determining a crisis score trend using the crisis score and the one or more historical crisis scores; wherein execution of the abatement protocol is further based on the crisis score trend.
20. The computer-program product of claim 17 , wherein the operations further include:
obtaining additional training data, the additional training data including one or more additional input phrases previously supplied to the trained machine learning classifier, each of the one or more additional input phrases being associated with a respective label indicative of whether a user confirmed or denied a respective crisis situation after the respective additional input phrase triggered the trained machine learning classifier, and
further training the trained machine learning classifier based at least in part on the additional training data, wherein further training occurs prior to receiving the input phrase.
21. The computer-program product of claim 20 , wherein each of the additional input phrases was previously supplied to the trained machine learning classifier and triggered the trained machine learning classifier.
22. The computer-program product of claim 17 , wherein executing the abatement protocol includes i) accessing and presenting a text string containing emergency contact information, ii) accessing and presenting a link configured to begin initiation of an emergency contact connection upon actuation; iii) sending a signal to automatically begin initiation of an emergency contact connection; iv) initiating a patient health questionnaire protocol; v) generating a prompt for selecting a crisis management tool; vi) automatically initiating a crisis management tool; or vii) any combination of i-vi.
23. The computer-program product of claim 17 , wherein receiving the input phrase occurs via a text-based communication protocol, wherein executing the abatement protocol includes automatically initiating a crisis management tool via the text-based communication protocol, the crisis management tool including a workflow of prompts and/or comments for mitigating the crisis situation.
24. The computer-program product of claim 23 , wherein the crisis management tool, when initiated, facilitates:
providing, via the text-based communication protocol, a series of questions from a patient health questionnaire; and
receiving, via the text-based communication protocol, a response for each question of the series of questions.