Harnessing GPT-3.5 for text parsing in solid-state synthesis – case study of ternary chalcogenides

Thway, Maung; Low, Andre K. Y.; Khetan, Samyak; Dai, Haiwen; Recatala-Gomez, Jose; Chen, Andy Paul; Hippalgaonkar, Kedar

doi:10.1039/d3dd00202k

Cited by 7 publications

(3 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[1][2][3][4][5][6][7] These models have received substantial attention for the fact that they can be intuitively "programmed" or "taught" using daily conversational language, thereby assisting with diverse chemistry research tasks. [8][9][10][11][12][13][14][15][16][17][18][19][20] It is envisioned that the evolution from text-only to more dynamic, multi-modal LLMs will result in even more powerful and convenient AI assistants across various applications. 5,[21][22][23] The recent introduction of GPT-4V, with 'V' denoting its vision capability, stands as a testament to this progress.…”

Section: Introductionmentioning

confidence: 99%

Image and data mining in reticular chemistry powered by GPT-4V

Zheng,

He,

Khattab

et al. 2024

Digital Discovery

View full text Add to dashboard Cite

show abstract

Section: Introductionmentioning

confidence: 99%

Image and data mining in reticular chemistry powered by GPT-4V

Zheng,

He,

Khattab

et al. 2024

Digital Discovery

View full text Add to dashboard Cite

show abstract

“…The recent success of general-purpose large language models (LLMs) offers a new direction for chemical data extraction. − While LLMs can achieve notable performance on NER–REL extraction with task-specific prompt engineering obtaining an F1-score of ∼90% with GPT-4, their results can be inconsistent and highly dependent on the task and instructions provided. Fine-tuning LLMs for chemical tasks offers potential improvements, but this process demands large data sets of labeled examples and remains computationally intensive .…”

Section: Introductionmentioning

confidence: 99%

Versatile Deep Learning Pipeline for Transferable Chemical Data Extraction

Alshehri,

Horstmann,

You

2024

J. Chem. Inf. Model.

View full text Add to dashboard Cite

Chemical information disseminated in scientific documents offers an untapped potential for deep learning-assisted insights and breakthroughs. Automated extraction efforts have shifted from resource-intensive manual extraction toward applying machine learning methods to streamline chemical data extraction. While current extraction models and pipelines have ushered in notable efficiency improvements, they often exhibit modest performance, compromising the accuracy of predictive models trained on extracted data. Further, current chemical pipelines lack both transferability�where a model trained on one task can be adapted to another relevant task with limited examples�and extensibility, which enables seamless adaptability for new extraction tasks. Addressing these gaps, we present ChemREL, a versatile chemical data extraction pipeline emphasizing performance, transferability, and extensibility. ChemREL utilizes a custom, diverse data set of chemical documents, labeled through an active learning strategy to extract two properties: normal melting point and lethal dose 50 (LD 50 ). The normal melting point is selected for its prevalence in diverse contexts and wider literature, serving as the foundation for pipeline training. In contrast, LD 50 evaluates the pipeline's transferability to an unrelated property, underscoring variance in its biological nature, toxicological context, and units, among other differences. With pretraining and fine-tuning, our pipeline outperforms existing methods and GPT-4, achieving F1-scores of 96.1% for entity identification and 97.0% for relation mapping, culminating in an overall F1-score of 95.4%. More importantly, ChemREL displays high transferability, effectively transitioning from melting point extraction to LD 50 extraction with 10 randomly selected training documents. Released as an open-source package, ChemREL aims to broaden access to chemical data extraction, enabling the construction of expansive relational data sets that propel discovery.

show abstract

“…General purpose large language models (LLMs) are a form of generative artificial intelligence (AI), pretrained on a broad data set so they can be applied to many different tasks using natural language. Pretrained LLMs have been investigated for a wide variety of chemical tasks, , such as extracting structured data from the literature, − writing numerical simulation software, and education . LLM-based workflows have been used to plan syntheses of organic molecules , and metal–organic frameworks (MOFs). ,− Recent work has benchmarked materials science , and general chemical knowledge − of existing LLMs, and there are efforts to develop chemistry/materials-specific LLMs. , Fine-tuning LLMs on modest amounts of data improves performance for specific tasks, while still taking advantage of the general pretraining to provide basic symbol interpretation and output formatting guidance.…”

mentioning

confidence: 99%

Large Language Models for Inorganic Synthesis Predictions

Kim,

Jung,

Schrier

2024

J. Am. Chem. Soc.

View full text Add to dashboard Cite

We evaluate the effectiveness of pretrained and fine-tuned large language models (LLMs) for predicting the synthesizability of inorganic compounds and the selection of precursors needed to perform inorganic synthesis. The predictions of fine-tuned LLMs are comparable to�and sometimes better than�recent bespoke machine learning models for these tasks but require only minimal user expertise, cost, and time to develop. Therefore, this strategy can serve both as an effective and strong baseline for future machine learning studies of various chemical applications and as a practical tool for experimental chemists.

show abstract

Harnessing GPT-3.5 for text parsing in solid-state synthesis – case study of ternary chalcogenides

Abstract: In solid-state thermoelectrics, optimally doped single-phase compounds are necessary for advancing the state-of-the-art. Thermoelectric devices, which convert heat into electricity and vice-versa, rely on bulk materials, usually made using solid-state...

Cited by 7 publications

References 42 publications

Image and data mining in reticular chemistry powered by GPT-4V

Image and data mining in reticular chemistry powered by GPT-4V

Versatile Deep Learning Pipeline for Transferable Chemical Data Extraction

Large Language Models for Inorganic Synthesis Predictions

Contact Info

Product

Resources

About