Methods and systems for prompt evaluation-based query generation
AssigneeT-Mobile Innovations LLC
InventorsTimothy Indrieri, Guangrui Su
Life of this patent
Estimating the term…
Abstract
A method comprises training a data model using training data, wherein the training data comprises labelled datasets indicating an intent associated with a plurality of different predefined prompts, receiving a prompt from a user equipment (UE) operated by user, extracting one or more keywords from the prompt, wherein the one or more keywords are respectively associated with one or more categories, and wherein the one or more categories are respectively associated with one or more databases, generating a query based on the one or more keywords, wherein the query includes one or more parameters based on the one or more keywords; and querying the one or more databases based on the query to retrieve data requested in the prompt.
Claims20
A method for prompt evaluation-based query generation in a communication network, wherein the method comprises:
- training, by a data model application executing at a data processing system, a data model using training data, wherein the training data comprises labelled datasets indicating an intent associated with a plurality of different predefined prompts;
- receiving, by the data model application, a prompt from a user equipment operated by a user, wherein the prompt comprises text indicative of a request for data maintained across one or more databases in the communication network and communicatively coupled to the data processing system;
- determining, by the data model application using a probabilistic application, an intent parameter set based on the prompt, wherein the intent parameter set indicates one or more attributes of the data being requested in the prompt;
- parsing, by the data model application using a data extraction application running a Named Entity Recognition model, the prompt to extract one or more keywords associated with one or more categories based on the intent parameter set, wherein the one or more categories are respectively associated with one or more of the one or more databases in the communication network;
- generating, by the data model application using a query application, a query based on the intent parameter set and the one or more keywords, wherein a type of the query is based on the intent parameter set, and wherein the query includes one or more parameters based on the one or more keywords;
- querying, by the data model application using the query application, the one or more databases based on the query to retrieve the data requested in the prompt;
- and
- transmitting, by the data model application using the query application, a response based on the data to the user equipment.
A data processing system, comprising:
- a non-transitory memory;
- a processor communicatively coupled to the memory;
- and
- an application stored at the memory, which when executed by the processor, causes the processor to be configured to:
- train a data model using training data, wherein the training data comprises labelled datasets indicating an intent associated with a plurality of different predefined prompts;
- receive a prompt from a user equipment operated by a user;
- determine a score for each word in the prompt;
- determine an intent parameter set based on the score for each word in the prompt, wherein the intent parameter set indicates one or more attributes of data being requested in the prompt;
- extract one or more keywords from the prompt, wherein the one or more keywords are respectively associated with one or more categories, and wherein the one or more categories are respectively associated with one or more databases;
- generate a query based on the intent parameter set and the one or more keywords, wherein a type of the query is based on the intent parameter set, and wherein the query includes one or more parameters based on the one or more keywords;
- and
- query the one or more databases based on the query to retrieve data requested in the prompt.
A non-transitory computer-readable medium having instructions stored thereon that, when executed by a processor, cause the processor to:
- train a data model using training data, wherein the training data comprises labelled datasets indicating an intent associated with a plurality of different predefined prompts;
- receive a prompt from a user equipment operated by a user;
- determine a score for each word in the prompt;
- determine an intent parameter set based on the score for each word in the prompt, wherein the intent parameter set indicates one or more attributes of data being requested in the prompt;
- extract one or more keywords from the prompt, wherein the one or more keywords are respectively associated with one or more categories, and wherein the one or more categories are respectively associated with one or more databases;
- generate a query based on the intent parameter set and the one or more keywords, wherein a type of the query is based on the intent parameter set, and wherein the query includes one or more parameters based on the one or more keywords;
- and
- query the one or more databases based on the query to retrieve data requested in the prompt.
Dependencies are inferred from the claim text.