Systems and methods for artificial-intelligence-augmented data retrieval
A disclosed method may include receiving, via a user interface, a natural language query that describes a task. The method may also include classifying, via a classification machine learning model, the natural language query as a query type included in a predetermined plurality of query types. The method may also include generating, by the processor via a generative machine learning model pre-trained to generate responses to queries based on at least one predetermined task execution policy, a response based on the natural language query and the query type. The method may also include providing, via the user interface, the response. Various other methods, systems, and computer-readable media are also disclosed.
1 . A method comprising:
receiving, by a processor, via a user interface, a natural language query that describes a task;
classifying, by the processor, via a classification machine learning model, the natural language query as a query type included in a predetermined plurality of query types;
selecting, based on the query type, an operation mode for a generative machine learning model from a plurality of operation modes by one of selecting a routing operation mode when the natural language query is classified as a routing query type or selecting a query-response operation mode when the natural language query is classified as a general query type;
generating, by the processor via the generative machine learning model pre-trained to generate responses to queries based on at least one predetermined task execution policy, a response based on the natural language query and the query type; and
providing, by the processor via the user interface, the response.
2 . The method of claim 1 , further comprising pre-training the classification machine learning model to classify natural language queries into one of a predetermined plurality of query types.
3 . The method of claim 1 , further comprising classifying the natural language query as the query type by classifying the natural language query as at least one of:
a routing query type; or
a general query type.
4 . The method of claim 1 , further comprising, when in the routing operation mode, selecting, via the generative machine learning model and based on the natural language query, a task execution path from a policy routing table included in the at least one predetermined task execution policy.
5 . The method of claim 4 , further comprising dynamically updating the policy routing table based on received changes to the at least one predetermined task execution policy.
6 . The method of claim 1 , further comprising, when in the query-response operation mode, via the generative machine learning model:
processing, utilizing a retrieval-augmented generation (RAG) approach, the at least one predetermined task execution policy to generate a processed at least one predetermined task execution policy;
storing the processed at least one predetermined task execution policy in a vector database; and
calculating similarity scores among entries in the vector database to find relevant policy sections to answer the natural language query.
7 . The method of claim 6 , further comprising updating the vector database when a new task execution policy is inputted.
8 . The method of claim 1 , further comprising deploying the generative machine learning model via a centralized inference service (CIS) and an internal large language model (LLM) platform.
9 . The method of claim 1 , further comprising evaluating, via the generative machine learning model, a likelihood of a selected action path resulting in a successful task execution based on the natural language query.
10 . The method of claim 1 , further comprising providing the response as at least one strategy selected from a group of strategies consisting of:
a task execution path,
a policy explanation, and
a prediction of a user's task execution requirements.
11 . The method of claim 1 , further comprising configuring the user interface to receive a policy upload to provide context for the generative machine learning model.
12 . The method of claim 1 , further comprising further basing generating the response on a user profile.
13 . The method of claim 1 , further comprising providing a supervisory feedback interface that enables a user to provide supervisory feedback to the generative machine learning model.
14 . A non-transitory computer-readable storage medium for tangibly storing computer program instructions capable of being executed by a computer processor, the computer program instructions defining steps of:
receiving, by a processor, via a user interface, a natural language query that describes a task;
classifying, by the processor, via a classification machine learning model, the natural language query as a query type included in a predetermined plurality of query types;
selecting, based on the query type, an operation mode for a generative machine learning model from a plurality of operation modes by one of selecting a routing operation mode when the natural language query is classified as a routing query type or selecting a query-response operation mode when the natural language query is classified as a general query type;
generating, by the processor via the generative machine learning model pre-trained to generate responses to queries based on at least one predetermined task execution policy, a response based on the natural language query and the query type; and
providing, by the processor via the user interface, the response.
15 . The non-transitory computer-readable storage medium of claim 14 , the steps further comprising pre-training the classification machine learning model to classify natural language queries into one of a predetermined plurality of query types.
16 . A device comprising:
a processor; and
a non-transitory computer-readable medium that, when executed by the processor, performs the steps of
receiving via a user interface, a natural language query that describes a task;
classifying via a classification machine learning model, the natural language query as a query type included in a predetermined plurality of query types;
selecting, based on the query type, an operation mode for a generative machine learning model from a plurality of operation modes by one of selecting a routing operation mode when the natural language query is classified as a routing query type or selecting a query-response operation mode when the natural language query is classified as a general query type;
generating, via the generative machine learning model pre-trained to generate responses to queries based on at least one predetermined task execution policy, a response based on the natural language query and the query type; and
providing, via the user interface, the response.
17 . The device of claim 16 , the steps further comprising pre-training the classification machine learning model to classify natural language queries into one of a predetermined plurality of query types.
18 . The device of claim 16 , the steps further comprising classifying the natural language query as the query type by classifying the natural language query as at least one of:
a routing query type; or
a general query type.
19 . The device of claim 16 , the steps further comprising evaluating, via the generative machine learning model, a likelihood of a selected action path resulting in a successful task execution based on the natural language query.
20 . The device of claim 16 , the steps further comprising providing the response as at least one strategy selected from a group of strategies consisting of:
a task execution path,
a policy explanation, and
a prediction of a user's task execution requirements.