This study introduces the use of large language models (LLMs), such as GPT, to accelerate materials language processing (MLP) tasks. MLP aims to extract structured data from research papers to support materials science research. Traditional MLP methods face challenges such as complex architectures, extensive fine-tuning, and reliance on large labeled datasets. This study proposes using GPT models for MLP tasks, leveraging prompt engineering to replace complex architectures. GPT models can perform text classification, named entity recognition (NER), and extractive question answering (QA) with limited datasets, demonstrating high performance in various materials science domains. These models can also identify incorrect annotated data, reducing the need for manual labeling. The study focuses on two key MLP tasks: text classification and information extraction. For text classification, the study presents a paper filtering method using zero-shot and few-shot learning, achieving high accuracy without fine-tuning. For information extraction, the study proposes an entity-centric prompt engineering method for NER, which outperforms previous fine-tuned models on multiple datasets. The study also introduces a GPT-enabled extractive QA model that provides precise answers to materials science questions. The study evaluates the performance of GPT models in text classification and NER tasks, comparing them with state-of-the-art models. Results show that GPT models achieve high accuracy and reliability, even with limited datasets. The study also demonstrates that GPT models can correct incorrect annotations in QA datasets, improving the quality of information extracted from materials science literature. The study concludes that GPT models offer a promising solution for MLP tasks, providing high performance and reliability while reducing the need for extensive manual labeling. However, the study also highlights potential limitations, such as the need for careful evaluation of results and the potential for overconfidence due to biases in training data. Despite these challenges, GPT models are expected to be effective tools for materials scientists, enabling more efficient analysis of literature without requiring expertise in complex NLP architectures.This study introduces the use of large language models (LLMs), such as GPT, to accelerate materials language processing (MLP) tasks. MLP aims to extract structured data from research papers to support materials science research. Traditional MLP methods face challenges such as complex architectures, extensive fine-tuning, and reliance on large labeled datasets. This study proposes using GPT models for MLP tasks, leveraging prompt engineering to replace complex architectures. GPT models can perform text classification, named entity recognition (NER), and extractive question answering (QA) with limited datasets, demonstrating high performance in various materials science domains. These models can also identify incorrect annotated data, reducing the need for manual labeling. The study focuses on two key MLP tasks: text classification and information extraction. For text classification, the study presents a paper filtering method using zero-shot and few-shot learning, achieving high accuracy without fine-tuning. For information extraction, the study proposes an entity-centric prompt engineering method for NER, which outperforms previous fine-tuned models on multiple datasets. The study also introduces a GPT-enabled extractive QA model that provides precise answers to materials science questions. The study evaluates the performance of GPT models in text classification and NER tasks, comparing them with state-of-the-art models. Results show that GPT models achieve high accuracy and reliability, even with limited datasets. The study also demonstrates that GPT models can correct incorrect annotations in QA datasets, improving the quality of information extracted from materials science literature. The study concludes that GPT models offer a promising solution for MLP tasks, providing high performance and reliability while reducing the need for extensive manual labeling. However, the study also highlights potential limitations, such as the need for careful evaluation of results and the potential for overconfidence due to biases in training data. Despite these challenges, GPT models are expected to be effective tools for materials scientists, enabling more efficient analysis of literature without requiring expertise in complex NLP architectures.