Extensible search, content, and dialog management system with human-in-the-loop curation
The present disclosure describes systems and methods for extensible search, content, and dialog management. Embodiments of the present disclosure provide a dialog system with a trained intent recognition model (e.g., a deep learning model) to receive and understand a natural language query from a user. In cases where intent is not identified for a received query, the dialog system generates one or more candidate responses that may be refined (e.g., using human-in-the-loop curation) to generate a response. The intent recognition model may be updated (e.g., retrained) the accordingly. Upon receiving a subsequent query with similar intent, the dialog system may identify the intent using the updated intent recognition model.
1. A method for dialog and search management, comprising:
receiving, via a dialog interface on a user device, a query from a user interaction;
determining, at a search management system operating on the user device, that an intent recognition model did not identify an intent tag for the query, wherein the intent recognition model is not trained to recognize the intent tag;
transmitting, by the search management system to a dialog curation system operating on a cloud server, query information indicating the determination to the dialog curation system;
receiving, by the search management system from the dialog curation system in response to transmitting the query information, an updated intent recognition model trained to recognize the intent tag;
replacing, by the search management system, the intent recognition model with the updated intent recognition model received from the dialog curation system;
receiving, via the dialog interface, a subsequent query from a subsequent user interaction;
determining, using the updated intent recognition model, that the subsequent query corresponds to the intent tag;
generating a response based on the intent tag; and
providing, to the dialog interface, the response for the subsequent query.
2. The method of claim 1 , further comprising:
determining that the query comprises a question.
3. The method of claim 1 , further comprising:
identifying a contextual intelligence framework (CIF) for the subsequent query, wherein the response is based at least in part on the CIF.
4. The method of claim 1 , further comprising:
transmitting the subsequent query to an offline search cache; and
receiving cached results for the subsequent query from the offline search cache, wherein the response is based at least in part on the cached results.
5. The method of claim 1 , further comprising:
responding that an answer to the query is unknown.
6. A method for dialog curation, comprising:
receiving, by a dialog curation system operating on a cloud server from a search management system operating on a user device, query information indicating that an intent recognition model did not identify an intent tag for a query, wherein the intent recognition model is not trained to recognize the intent tag;
generating a response to the query based on the query information;
updating the intent recognition model to include the intent tag corresponding to the query using a training model, wherein the updated intent recognition model is trained to recognize the intent tag;
replacing, by the search management system, the intent recognition model with the updated intent recognition model received from the dialog curation system; and
transmitting the response and the updated intent recognition model from the dialog curation system to the search management system.
7. The method of claim 6 , further comprising:
generating a candidate response using a dialog generation engine; and
refining the candidate response to generate the response.
8. The method of claim 7 , wherein:
the candidate response is refined by a human curator.
9. The method of claim 6 , further comprising:
linking an additional query with the response, wherein the intent recognition model is updated based at least in part on the additional query.
10. The method of claim 6 , further comprising:
receiving the query from the search management system, wherein the search management system receives the query from a user interaction and transmitting the query information to the dialog curation system after attempting to detect the intent tag of the query using the intent recognition model and determining that the intent recognition model did not identify the intent tag for the query.
11. The method of claim 6 , further comprising:
receiving a subsequent query; and
determining, using the updated intent recognition model, that the subsequent query corresponds to the updated intent tag.
12. The method of claim 6 , further comprising:
updating a content scripting engine based on the response.
13. A system for dialog management, comprising:
one or more processors; and
one or more memories including instructions executable by the one or more processors to:
generate, by a dialog generation engine on a dialog curation system operating on a cloud server, a response based on query information received from a search management system operating on a user device, wherein the query information indicates that an intent recognition model did not identify an intent tag for a query, and wherein the intent recognition model is not trained to recognize the intent tag; and
update, by a training model, the intent recognition model and a content scripting engine, wherein the updated intent recognition model is trained to recognize the intent tag;
replacing, by the search management system operating on the user device, the intent recognition model with the updated intent recognition model received from the dialog curation system.
14. The system of claim 13 , wherein:
the dialog generation engine comprises a neural network with a transformer architecture.
15. The system of claim 13 , wherein the one or more memories further including instructions executable by the one or more processors to receive, via a dialog interface, the query and to provide the response to a user.
16. The system of claim 15 , wherein:
the dialog interface comprises a plurality of sections corresponding to a plurality of content types.
17. The system of claim 13 , wherein the one or more memories further including instructions executable by the one or more processors to provide a contextual intelligence framework (CIF) for the query by a CIF component.
18. The system of claim 13 , wherein the one or more memories further including instructions executable by the one or more processors to determine whether the query comprises a question by a question detection model.
19. The system of claim 13 , wherein the one or more memories further including instructions executable by the one or more processors to provide offline search results based on the query by an offline search cache.
20. The system of claim 13 , wherein the one or more memories further including instructions executable by the one or more processors to refine the response generated by the dialog generation engine by a curation interface.