Accelerating agent performance in a natural language processing system
A computer-implemented method for providing agent assisted transcriptions of user utterances. A user utterance is received in response to a prompt provided to the user at a remote client device. An automatic transcription is generated from the utterance using a language model based upon an application or context, and presented to a human agent. The agent reviews the transcription and may replace at least a portion of the transcription with a corrected transcription. As the agent inputs the corrected transcription, accelerants are presented to the user comprising suggested texted to be inputted. The accelerants may be determined based upon an agent input, an application or context of the transcription, the portion of the transcription being replaced, or any combination thereof. In some cases, the user provides textual input, to which the agent transcribes an intent associated with the input with the aid of one or more accelerants.
1. A computer-implemented method comprising:
receiving over a network, at an application at an application server, an utterance provided by a user of a client device in response to a prompt presented to the user by the application;
determining a context of the associated utterance indicating an interaction state between the user and the application, based at least in part upon one or more previous prompts presented to the user;
generating at least one automatic transcription of the utterance using a language model selected from a plurality of language models based upon the determined context;
generating an agent interface to be displayed to an agent at an agent terminal, the agent different from the user of the remote client device, the agent interface displaying at least the at least one automatic transcription;
receiving, from the agent, via the agent interface displayed at the agent terminal, one or more inputs indicating replacement or supplement of at least a portion of the at least one automatic transcription;
in response to receiving the one or more inputs, presenting to the agent via the agent interface one or more suggestions based upon the one or more inputs, wherein the one or more suggestions are updated as additional inputs from the agent are received; and
generating an updated transcription of the utterance.
2. The method of claim 1 , wherein the context is determined based upon traversal of a context model using the one or more previous prompts presented to the user.
3. The method of claim 2 , wherein traversal of the context model is further based upon and one or more previous utterances received from the user responsive to presentation of the one or more previous prompts.
4. The method of claim 1 , further comprising:
generating a priming string based upon the presented prompt and the determined context, the priming string describing an expected indication associated with the utterance to aid an agent in transcribing the utterance; and
wherein the agent interface further displays the generated priming string.
5. The method of claim 4 , wherein the priming string describes an expected text type associated with the utterance.
6. The method of claim 4 , wherein the generated priming string is different from a second priming string generated based upon a same prompt presented to a second user of the application associated with a different context.
7. The method of claim 1 , wherein the language model is selected from the plurality of language models comprising:
determining whether the plurality of language models comprises a language model specific to the identified context and the application;
responsive to determining that the plurality of language models does not comprise a language model specific to the identified context and application, selecting a language model of the plurality of language models corresponding to the application.
8. The method of claim 1 , further comprising updating the selected language model based upon the updated transcription.
9. The method of claim 1 , wherein the one or more inputs comprises a piece of text for replacing or supplementing the portion of the automatic transcription.
10. The method of claim 9 , wherein the one or more suggestions are selected such that at least a portion of each of the one or more suggestions matches the piece of text.
11. The method of claim 1 , wherein the one or more suggestions are generated using a second language model selected based upon the determined context.
12. The method of claim 1 , further comprising:
determining an agent speed based upon a number of actions taken by the agent to generate the updated transcription;
determining a potential speed based upon a potential number of actions needed to generate the updated transcription using the one or more suggestions;
responsive to the agent speed and the potential speed differing by at least a threshold amount, generating a notification for display to the agent at the agent terminal.
13. A language processing system comprising a processor and a memory in communication with the processor, the memory storing programming instructions executable by the processor to:
receive over a network, at an application at an application server, an utterance provided by a user of a client device in response to a prompt presented to the user by the application;
determine a context of the associated utterance indicating an interaction state between the user and the application, based at least in part upon one or more previous prompts presented to the user;
generate at least one automatic transcription of the utterance using a language model selected from a plurality of language models based upon the determined context;
generate an agent interface to be displayed to an agent at an agent terminal, the agent different from the user of the remote client device, the agent interface displaying at least the at least one automatic transcription;
receive, from the agent, via the agent interface displayed at the agent terminal, one or more inputs indicating replacement or supplement of at least a portion of the at least one automatic transcription;
in response to receiving the one or more inputs, present to the agent via the agent interface one or more suggestions based upon the one or more inputs, wherein the one or more suggestions are updated as additional inputs from the agent are received; and
generate an updated transcription of the utterance.
14. The language processing system of claim 13 , wherein the context is determined based upon traversal of a context model using the one or more previous prompts presented to the user.
15. The language processing system of claim 13 , wherein the programming instructions are further executable by the processor to:
generating a priming string based upon the presented prompt and the determined context, the priming string describing an expected indication associated with the utterance to aid an agent in transcribing the utterance; and
wherein the agent interface further displays the generated priming string.
16. The language processing system of claim 15 , wherein the generated priming string is different from a second priming string generated based upon a same prompt presented to a second user of the application associated with a different context.
17. The language processing system of claim 13 , wherein the language model is selected from the plurality of language models comprising:
determining whether the plurality of language models comprises a language model specific to the identified context and the application;
responsive to determining that the plurality of language models does not comprise a language model specific to the identified context and application, selecting a language model of the plurality of language models corresponding to the application.
18. The language processing system of claim 13 , wherein the one or more inputs comprises a piece of text for replacing or supplementing the portion of the automatic transcription, and the one or more suggestions are selected such that at least a portion of each of the one or more suggestions matches the piece of text.
19. The language processing system of claim 13 , wherein the one or more suggestions are generated using a second language model selected based upon the determined context.
20. The language processing system of claim 13 , wherein the programming instructions are further executable by the processor to:
determine an agent speed based upon a number of actions taken by the agent to generate the updated transcription;
determine a potential speed based upon a potential number of actions needed to generate the updated transcription using the one or more suggestions;
responsive to the agent speed and the potential speed differing by at least a threshold amount, generate a notification for display to the agent at the agent terminal.