Enhanced language model for processing user-specific queries

A system comprising:

• a computer processor;
• a data repository in communication with the computer processor and storing:
• a user-specific query relating to a user,
• user data from a user-specific data source associated with the user and related to the user-specific query,
• a graph data structure that defines, as edges, relationships between data elements contained in the user data,
• wherein the data elements comprise nodes of the graph data structure,
• wherein the nodes are connected by edges, and
• wherein the graph data structure is stored in a data schema, a prompt,
• a context comprising context data stored in the data schema,
• wherein the context is at least partially related to the user-specific query, and
• wherein the data schema for the graph data structure and the data schema for the context data are a same data schema format, and
• an output;
• a language model executable by the computer processor, wherein the language model is trained with the context;
• and
• a server controller programmed, when executed by the computer processor, to perform a computer-implemented method comprising:
• receiving the user-specific query,
• extracting the user data,
• converting the user data into the graph data structure,
• constructing the prompt, wherein constructing comprises:
• adding, to the prompt, the graph data structure such that, when executed, the language model processes both the edges and the nodes in the graph data structure,
• adding, to the prompt, a first command to reference the context, and
• adding, to the prompt, a second command to generate the output from the context, wherein the output is indirectly related to the user data, and
• executing the language model with the prompt to generate the output.

A method of training a language model, the method comprising:

• extracting user data relating to a user from a user-specific data source associated with the user and related to a user-specific query;
• converting the user data into a graph data structure that defines, as edges, relationships between data elements contained in the user data,
• wherein the data elements comprise nodes of the graph data structure,
• wherein the nodes are connected by the edges, and
• wherein the graph data structure is stored in a data schema;
• constructing a prompt for the language model, wherein constructing comprises:
• adding, to the prompt, the graph data structure such that, when executed, the language model processes both the edges and the nodes in the graph data structure,
• adding, to the prompt, a first command to reference a context comprising context data stored in the data schema,
• wherein the context is at least partially related to the user-specific query, and
• wherein the data schema for the graph data structure and the data schema for the context data are a same data schema format, and
• adding, to the prompt, a second command to generate an output from the context, wherein the output is indirectly related to the user data;
• executing the language model with the prompt to generate the output, wherein the language model is trained with the context;
• returning the output to a data repository also storing training data;
• and
• retraining the language model on the output and the training data.