Conversation system for detecting a dangerous mental or physical condition
The present disclosure describes a system to use conversation data of patients to detect dangerous mental or physical conditions, such as suicidal thoughts, physical abuse, recent falls, and viral infection. A machine learning system may be trained to identify a dangerous mental or physical condition from conversations based on examples of patients evaluated to have a specific mental or physical condition. Conversations of patients may be monitored, natural language understanding (NLU) processing performed, and a machine learning system used to detect dangerous mental or physical conditions.
1. A computer-implemented method, comprising:
receiving conversation data for an individual senior patient whose conversations are monitored with at least one other peer senior;
generating a conversation feature vector from the conversation data, the conversation feature vector including conversation topics of the individual senior patient, conversation entities mentioned by the individual senior patient, and conversation sentiment;
providing the conversation feature vector as an input for a classifier trained at least in part on training data that includes conversation data for a group of senior patients evaluated to have a dangerous mental or physical condition including at least one member from the group consisting of: suicidal ideas, physical falls, physical abuse, and a viral infection; and
identifying, from an output of the classifier, whether or not the individual senior patient is at risk of having the dangerous physical or mental condition.
2. The computer-implemented method of claim 1 , further comprising extracting feature vectors from the conversation data for the individual senior patient and applying the feature vectors to a neural network classifier built based at least in part on training data for people in the group of senior patients who have had the dangerous mental or physical condition.
3. The computer-implemented method of claim 1 , wherein the classifier comprises a neural network classifier, and the method further comprises, training the neural network classifier to generate a decision whether the individual senior patient has the dangerous mental or physical condition.
4. The computer-implemented method of claim 3 , further comprising building a neural network classifier for a specific mental or physical condition by selecting training sets, testing sets, and evaluation sets based on data for a set of senior patients having the dangerous mental or physical condition.
5. The computer-implemented method of claim 1 , further comprising generating an alert in response to detecting that the individual senior patient is at risk of having the dangerous mental or physical condition.
6. The computer-implemented method of claim 1 , further comprising generating a user interface indicating whether the individual senior patient is at risk of having the dangerous physical or mental condition.
7. The computer-implemented method of claim 1 , wherein the conversation data for the individual senior patient is generated by monitoring conversations, performing natural language processing on monitored conversations, and generating conversation objects.
8. A computer-implemented method comprising:
receiving conversation data for a group of senior patients whose conversations with at least one senior are monitored;
for an individual senior patient, constructing a conversation feature vector, from conversation data associated with the individual senior patient, for a dangerous physical or mental condition including at least one member from the group consisting of: suicidal ideas, physical falls, physical abuse, and a viral infection;
applying the conversation feature vector to a neural network classifier trained to identify a likelihood that a senior patient has the dangerous physical or mental condition, wherein the neural network classifier is built based at least in part from training data based on a set of senior patients, from the group of senior patients, evaluated as having the dangerous physical or mental condition; and
generating an output indicating whether or not the individual senior patient is at risk of having the dangerous physical or mental condition.
9. The computer-implemented method of claim 8 , further comprising generating a user interface indicating whether the individual senior patient is at risk of having the dangerous physical or mental condition.
10. The computer-implemented method of claim 8 , wherein the conversation feature vector includes conversation topics of the individual senior patient, conversation entities mentioned by the individual senior patient, and conversation sentiment.
11. A system comprising:
a processor; and
a memory storing one or more instructions that, when executed, cause the processor to implement operations including:
receiving conversation data for an individual senior patient whose conversations are monitored with at least one other peer senior;
conversation feature vector including conversation topics of the individual senior patient, conversation entities mentioned by the individual senior patient, and conversation sentiment;
providing the conversation feature vector as an input for a classifier trained at least in part on training data that includes conversation data for a group of senior patients evaluated to have a dangerous mental or physical condition including at least one member from the group consisting of: suicidal ideas, physical falls, physical abuse, and a viral infection; and
identifying, from an output of the classifier, whether or not the individual senior patient is at risk of having the dangerous physical or mental condition.
12. The system of claim 11 , wherein the operations include extracting feature vectors from the conversation data for the individual senior patient and extracting the feature vectors to a neural network classifier built based at least in part on training data for people in the group of senior patients who have had the dangerous mental or physical condition.
13. The system of claim 11 , wherein the operations further comprising building a neural network classifier for a specific mental or physical condition by selecting training sets, testing sets, and evaluation sets based on data for a set of senior patients evaluated for the dangerous mental or physical condition.
14. The system of claim 11 , wherein the operations comprise constructing feature vectors for conversations based on a set of conversation criteria associated with the dangerous mental or physical condition.
15. The system of claim 11 , wherein the operation further comprise generating an alert when the machine learning system detects that the individual senior patient has the dangerous mental or physical condition.
16. The system of claim 11 , wherein the operations further comprise generating a user interface indicating whether the individual senior patient is at risk of having the dangerous physical or mental condition.
17. The system of claim 15 , wherein the conversation data for the individual senior patient is generated by monitoring patient conversations in living areas of senior patients, performing natural language processing on monitored conversations, and generating conversation objects.
18. A system comprising:
a processor; and
a memory storing one or more instructions that, when executed, cause the processor to implement operations including:
receiving conversation data for a group of senior patients whose conversations with an individal senior patient are monitored;
constructing a conversation feature vector, from conversation data associated with the individual senior patient, for a dangerous physical or mental condition including at least one member from the group consisting of: suicidal ideas, physical falls, physical abuse, and a viral infection;
applying the conversation feature vector to a neural network classifier trained to identify a likelihood that a senior patient has the dangerous physical or mental condition, wherein the neural network classifier is built based at least in part from training data based on a set of senior patients, from the group of senior patients, evaluated as having the dangerous physical or mental condition; and
generating an output indicating whether or not the individual senior patient is at risk of having the dangerous physical or mental condition.