Dynamic language generation for explainable outputs in conversational language models

Abstract

In some embodiments, a system may enable autonomous reasoning in real-time communications by dynamically generating and executing DSLs based on context. The method includes determining a computational intent specification from messages exchanged in a communication session and accessing knowledge sources relevant to the computational intent specification. The method further includes generating a DSL using a generative language model based on the computational intent, the messages, and the knowledge sources, where the language defines constructs and syntax rules specifying allowable compositions of named entities and relationships. The method further includes executing the constructs to perform reasoning operations and generate an output message responsive to the messages. The method further includes presenting the output message within the real-time communication platform and storing an execution trace mapping the constructs to portions of the output message.