Automated workflow selection for agent within an electronic conversation system
Within an environment in which users converse at least partly with human agents to accomplish a desired task, a server assists the agents by identifying workflows that are most applicable to the current conversation. Workflow selection functionality identifies one or more candidate workflows based on techniques such as user intent inference, conversation state tracking, or search, according to various embodiments. The identified candidate workflows are either automatically selected on behalf of the agent, or are presented to the agent for manual selection.
1. A computer-implemented method for assisting a human agent in a conversation between a human user and a system, the method comprising:
based on the conversation, predicting a plurality of alternative candidate next tasks to be performed for the user;
causing display, within a user interface of a human agent associated with the system, of indicators of a set of workflows corresponding to the plurality of alternative candidate next tasks, each workflow specifying a set of data items to obtain from the user;
identifying, from among the set of workflows, a workflow selected by the human agent;
causing display, within the user interface of the human agent, of a set of user interface elements for obtaining the set of data items specified by the identified workflow; and
refraining from causing display of sets of user interface elements for obtaining the sets of data items of workflows of the set of workflows other than the workflow selected by the human agent.
2. The computer-implemented method of claim 1 , further comprising using the identification of the workflow selected by the human agent, and prior words of the conversation, to train a machine learning model that takes a conversation as input and outputs a set of one or more candidate workflow identifiers.
3. The computer-implemented method of claim 2 , wherein the predicting of the plurality of candidate tasks is performed using the machine learning model.
4. The computer-implemented method of claim 1 , wherein the predicting of the plurality of candidate tasks comprises:
tracking conversation state of the conversation; and
predicting the plurality of candidate tasks according to a probability distribution based on the state.
5. The computer-implemented method of claim 4 , wherein the tracking of conversation state uses at least one of: conversation intent, conversation entities, or conversation words.
6. The computer-implemented method of claim 1 , wherein the predicting of the plurality of candidate tasks comprises applying preexisting regular expressions corresponding to the plurality of candidate tasks.
7. The computer-implemented method of claim 1 , wherein the predicting of the plurality of candidate tasks comprises searching an index of workflows indexed by words in the user interface elements.
8. The computer-implemented method of claim 1 , further comprising ingesting a new workflow into a workflow repository, the ingesting comprising indexing textual metadata associated with the new workflow.
9. The computer-implemented method of claim 1 , wherein the predicting of the plurality of candidate tasks is performed at each turn of the conversation.
10. A non-transitory computer-readable medium storing instructions that when executed by a computer processor perform actions comprising:
based on a conversation between a human user and a system, predicting a plurality of alternative candidate next tasks to be performed for the user;
causing display, within a user interface of a human agent associated with the system, of indicators of a set of workflows corresponding to the plurality of alternative candidate next tasks, each workflow specifying a set of data items to obtain from the user;
identifying, among the set of workflows, a workflow selected by the human agent;
causing display, within the user interface of the human agent, of a set of user interface elements for obtaining the set of data items specified by the identified workflow; and
refraining from causing display of sets of user interface elements for obtaining the sets of data items of workflows of the set of workflows other than the workflow selected by the human agent.
11. The non-transitory computer-readable medium of claim 10 , the actions further comprising using the identification of the workflow selected by the human agent, and prior words of the conversation, to train a machine learning model that takes a conversation as input and outputs a set of one or more candidate workflow identifiers.
12. The non-transitory computer-readable medium of claim 11 , wherein the predicting of the plurality of candidate tasks is performed using the machine learning model.
13. The non-transitory computer-readable medium of claim 10 , wherein the predicting of the plurality of candidate tasks comprises:
tracking conversation state of the conversation; and
predicting the plurality of candidate tasks according to a probability distribution based on the state.
14. The non-transitory computer-readable medium of claim 13 , wherein the tracking of conversation state uses at least one of: conversation intent, conversation entities, or conversation words.
15. The non-transitory computer-readable medium of claim 10 , wherein the predicting of the plurality of candidate tasks comprises applying preexisting regular expressions corresponding to the plurality of candidate tasks.
16. The non-transitory computer-readable medium of claim 10 , wherein the predicting of the plurality of candidate tasks comprises searching an index of workflows indexed by words in the user omterface elements.
17. The non-transitory computer-readable medium of claim 10 , the actions further comprising ingesting a new workflow into a workflow repository, the ingesting comprising indexing textual metadata associated with the new workflow.
18. The non-transitory computer-readable medium of claim 10 , wherein the predicting of the plurality of candidate tasks is performed at each turn of the conversation.
19. A computer server comprising:
a computer processor; and
a non-transitory computer-readable medium storing instructions that when executed by the computer processor perform actions comprising:
based on a conversation of a human user, predicting a plurality of alternative candidate next tasks to be performed for the user;
causing display, within a user interface of a human agent associated with the computer server, of indicators of a set of workflows corresponding to the plurality of alternative candidate next tasks, each workflow specifying a set of data items to obtain from the user;
identifying, among the set of workflows, a workflow selected by the human agent;
causing display, within the user interface of the human agent, of a set of user interface elements for obtaining the set of data items specified by the identified workflow; and
refraining from causing display of sets of user interface elements for obtaining the sets of data items of workflows of the set of workflows other than the workflow selected by the human agent.
20. The computer server of claim 19 , the actions further comprising using the identification of the workflow selected by the human agent, and prior words of the conversation, to train a machine learning model that takes a conversation as input and outputs a set of one or more candidate workflow identifiers.