Automated pipeline for training language models

Abstract

The disclosed embodiments describe a method, system, and computer-readable medium for generating a training dataset for training a model in the field of natural language processing involving receiving a set of input samples and performing a rephrasing operation to produce new versions of the set of input samples, where the new versions preserve semantic equivalence as the set of input samples but have different phrasing. A dataset of generated versions of the input samples is generated using a generative Language Learning Model (LLM), all entity references present in the generated versions of the input samples are labeled, and the generated versions of the input samples and their corresponding labeled versions to form an expanded labeled dataset are aggregated.

Dependencies are inferred from the claim text.