Process mining asynchronous support conversation using attributed directly follows graphing
Provided is a computer-implemented method, system, and computer program product for process mining asynchronous support conversations using attributed directly follows graphing. A processor may collect a plurality of conversation threads from an asynchronous data stream. The processor may label each utterance of a plurality of utterances from the plurality of conversation threads with an event label. The processor may analyze the event label for each utterance of the plurality of utterances. The processor may generate, based on the analyzing of the event label for each utterance, an attributed directly follows graph (DFG).
1. A computer-implemented method comprising:
collecting a plurality of conversation threads related to a task from an asynchronous data stream;
labeling each utterance of a plurality of utterances from the plurality of conversation threads with an event label;
analyzing the event labels for each utterance of the plurality of utterances;
generating, based on the analyzing, an attributed directly follows graph (DFG), wherein the DFG transforms the plurality of conversation threads into domain-specific, attributed-activity trace instances with rationalized edges and nodes, and wherein the DFG includes a minimized edge and node frequency based on an edge/node rationalization;
identifying a cause of a process bottleneck or task failure in one or more digital assistants based on the attributed DFG; and
modifying one or more algorithms in the one or more digital assistants that resolve the cause; and
completing the task using the one or more modified algorithms in the one or more digital assistants.
2. The computer-implemented method of claim 1 , wherein analyzing the event labels for each utterance of the plurality of utterances further comprises:
generating, for each thread of the plurality of conversation threads, a role-specific ordered list of users within the thread; and
classifying each utterance of the plurality utterances that are associated with each user of each thread to identify a type of utterance.
3. The computer-implemented method of claim 2 , wherein classifying each utterance of the plurality of utterances that are associated with each user to identify the type of utterance comprises:
classifying a given utterance of the plurality of utterances as the type selected from a group of utterance types consisting of an acknowledgement, a status query, an information query, an action, and a handover.
4. The computer-implemented method of claim 3 , further comprising:
filtering one or more utterances that failed to be classified.
5. The computer-implemented method of claim 2 , further comprising:
merging two or more utterances of the plurality of utterances that have a temporal proximity from a same user within the thread.
6. The computer-implemented method of claim 2 , further comprising:
marking, based on the event labels, a classification transition beyond a time proximity threshold as a turn boundary in a conversation thread.
7. The computer-implemented method of claim 1 , further comprising:
analyzing the attributed DFG to identify one or more incomplete conversations within the plurality of conversation threads; and
identifying, based on the analyzing, a digital assistant has failed to complete a task associated with a conversation thread.
8. The computer-implemented method of claim 7 , further comprising:
evaluating the digital assistant that has failed to complete the task; and
updating the digital assistant in order to complete the task.
9. A system comprising:
a processor; and
a computer-readable storage medium communicatively coupled to the processor and storing program instructions which, when executed by the processor, cause the processor to perform a method comprising:
collecting a plurality of conversation threads related to a task from an asynchronous data stream;
labeling each utterance of a plurality of utterances from the plurality of conversation threads with an event label;
analyzing the event labels for each utterance of the plurality of utterances;
generating, based on the analyzing, an attributed directly follows graph (DFG), wherein the DFG transforms the plurality of conversation threads into domain-specific, attributed-activity trace instances with rationalized edges and nodes, and wherein the DFG includes a minimized edge and node frequency based on an edge/node rationalization;
identifying a cause of a process bottleneck or task failure in one or more digital assistants based on the attributed DFG; and
modifying one or more algorithms in the one or more digital assistants that resolve the cause; and
completing the task using the one or more modified algorithms in the one or more digital assistants.
10. The system of claim 9 , wherein analyzing the event labels for each utterance of the plurality of utterances:
generating, for each thread of the plurality of conversation threads, a role-specific ordered list of users within the thread; and
classifying each utterance of the plurality utterances that are associated with each user of each thread to identify a type of utterance.
11. The system of claim 10 , wherein classifying each utterance of the plurality of utterances that are associated with each user to identify the type of utterance comprises:
classifying a given utterance of the plurality of utterances as the type selected from a group of utterance types consisting of an acknowledgement, a status query, an information query, an action, and a handover.
12. The system of claim 11 , wherein the method performed by the processor further comprises:
filtering one or more utterances that failed to be classified.
13. The system of claim 10 , wherein the method performed by the processor further comprises:
merging two or more utterances of the plurality of utterances that have a temporal proximity from a same user within the thread.
14. The system of claim 9 , wherein the method performed by the processor further comprises:
analyzing the attributed DFG to identify one or more incomplete conversations within the plurality of conversation threads; and
identifying, based on the analyzing, a digital assistant has failed to complete a task associated with a conversation thread.
15. A computer program product comprising a computer-readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform a method comprising:
collecting a plurality of conversation threads related to a task from an asynchronous data stream;
labeling each utterance of a plurality of utterances from the plurality of conversation threads with an event label;
analyzing the event labels for each utterance of the plurality of utterances;
generating, based on the analyzing, an attributed directly follows graph (DFG), wherein the DFG transforms the plurality of conversation threads into domain-specific, attributed-activity trace instances with rationalized edges and nodes, and wherein the DFG includes a minimized edge and node frequency based on an edge/node rationalization;
identifying a cause of a process bottleneck or task failure in one or more digital assistants based on the attributed DFG; and
modifying one or more algorithms in the one or more digital assistants that resolve the cause; and
completing the task using the one or more modified algorithms in the one or more digital assistants.
16. The computer program product of claim 15 , wherein analyzing the event labels for each utterance of the plurality of utterances:
generating, for each thread of the plurality of conversation threads, a role-specific ordered list of users within the thread; and
classifying each utterance of the plurality utterances that are associated with each user of each thread to identify a type of utterance.
17. The computer program product of claim 16 , wherein the method performed by the processor further comprises:
classifying a given utterance of the plurality of utterances as the type selected from a group of utterance types consisting of an acknowledgement, a status query, an information query, an action, and a handover; and
filtering one or more utterances that failed to be classified.
18. The computer program product of claim 15 , wherein the method performed by the processor further comprises:
merging two or more utterances of the plurality of utterances that have a temporal proximity from a same user within the thread.
19. The computer program product of claim 15 , wherein the method performed by the processor further comprises:
analyzing the attributed DFG to identify one or more incomplete conversations within the plurality of conversation threads; and
identifying, based on the analyzing, a digital assistant has failed to complete a task associated with a conversation thread.
20. The computer program product of claim 19 , wherein the method performed by the processor further comprises:
evaluating the digital assistant that has failed to complete the task; and
updating the digital assistant in order to complete the task.