Extracting ORR Catalyst Information for Fuel Cell from Scientific Literature

Amgad Ahmed Ali Ibrahim; Hein Htet; Ryoji Asahi; Yutaka Sasaki

REVIEW 4 major objections 6 minor 29 references

Extracting ORR Catalyst Information for Fuel Cell from Scientific Literature

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Fine-tuned BERT extracts fuel-cell catalyst data at 82% F1

desk verdict A genuinely new annotated ORR/PEFC catalyst corpus whose headline F1 numbers should be read as optimistic, because the test set and the Gold Standard are not clearly independent of training. read the letter →

arxiv 2507.07499 v1 pith:5ML6XCPH submitted 2025-07-10 cs.CL physics.data-an

Hein Htet , Amgad Ahmed Ali Ibrahim , Yutaka Sasaki , Ryoji Asahi This is my paper · ORCID

classification cs.CLphysics.data-an

keywords oxygenreductionreactionfuelcellcatalystsnamedentityrecognitionrelationextractionBERTfine-tuningmaterialsinformaticscorpusannotationDyGIE++

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that a carefully annotated corpus of oxygen reduction reaction (ORR) catalyst literature, combined with fine-tuned BERT models inside the DyGIE++ framework, can extract structured catalyst information from full-text articles at a level close to that of human annotators. It reports a named entity recognition F1-score of 82.19% and a relation extraction F1-score of 66.10% on an expert-verified Gold Standard, with PubMedBERT and SciBERT leading on entities and MatSciBERT on relations. The larger point is that automated literature mining of this kind could make catalyst databases buildable at scale, reducing the manual review burden in fuel cell research. The authors also assert that domain-specific BERT variants outperform the general biomedical model BlueBERT on this task.

What carries the argument

The machinery is DyGIE++, a joint named entity recognition and relation extraction framework that represents text spans in a dynamic graph, fine-tuned with seven BERT variants (SciBERT, three MatSciBERT versions, PubMedBERT, BlueBERT, BioBERT) on the FC-CoMIcs corpus. The corpus schema — 12 entity types (catalyst, support, additive, electrolyte, precursors, other material, material reference, property, structure, process, condition, value) and two relations (equivalent, related to) — defines what counts as a correct extraction. The pipeline also depends on ChemDataExtractor's parser for generating initial candidate annotations and the Brat tool for manual refinement.

What would settle it

An independent team annotating a held-out set of ORR catalyst articles with the same 12 entity types and 2 relations, with the current models evaluated only on that set; if NER F1 falls well below 82% (or RE below 66%), the reported scores overstate generalization.

Watch

Extended reading notes

Core claim

The central claim is that a manually annotated fuel cell corpus, FC-CoMIcs, built from 76 full-text articles (554 documents, 16,301 entities, 13,899 relations across 12 entity types and two relation types), is sufficient to train joint NER/RE models that extract ORR catalyst information nearly as well as trained human annotators. On the expert Gold Standard, the fine-tuned PubMedBERT and SciBERT models reach an NER F1 of 82.19% and MatSciBERT-1 reaches an RE F1 of 66.10%. When compared on the same article, the models' NER performance approached the annotators', whereas relation extraction still showed a noticeable gap. The authors take this as evidence that automated, scalable literature analysis for fuel cell catalysts is feasible.

Load-bearing premise

The reported F1 scores assume the Gold Standard is an independent test set, but it was created by the same expert group that trained the annotators whose labels form the training data, so the models may have indirectly learned Gold Standard annotation patterns.

Editorial extensions

If this is right

Domain-specific pre-trained models (SciBERT, MatSciBERT, PubMedBERT) outperform the general biomedical BlueBERT on ORR catalyst extraction, suggesting that matching pretraining domain to the corpus matters.
Joint NER/RE scores on the expert Gold Standard are high enough (82.19% NER, 66.10% RE) to support building structured catalyst databases from literature automatically.
The web-based annotation and extraction platform enables researchers to collect, annotate, and extract data without local installation, lowering the barrier for corpus expansion.
Because the models generalize from only 76 articles, scaling the corpus to more sources and more fuel cell technologies should further improve extraction coverage.
The released CSV dataset provides a structured view of catalyst–property–value relationships that can feed materials informatics pipelines.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A testable extension would be to retrain on an independently annotated test set produced by a different group; if the Gold Standard F1 drops sharply, the reported scores partly reflect annotation-pattern leakage rather than pure generalization.
The relation extraction gap (66% vs. human performance) suggests that the next bottleneck is not finding entities but resolving long-range and cross-sentence relations, so models that explicitly handle coreference or document-level context could yield the largest gains.
Because the corpus schema uses generic 'value' and 'condition' entities, the extracted data could be converted into quantitative catalyst benchmark tables (e.g., mass activity vs. synthesis conditions) that directly support property prediction models.
The same annotation-plus-fine-tuning recipe could transfer to other energy materials (e.g., electrolyzer catalysts or battery electrolytes) with modest additional annotation effort.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: this is a dataset paper, and the dataset is the contribution. The pipeline (DyGIE++ plus BERT variants) is borrowed from earlier materials IE work, but the annotated ORR/PEFC catalyst corpus — 76 articles, 554 documents, 16,301 entities, 13,899 relations — is new and plausibly useful for materials informatics. The authors also built a web-based annotation and extraction system, and they are honest about limitations, including the small dataset and cross-sentence relation failures.

What I trust: the corpus construction is described in enough detail to be reproduced, the entity/relation schema is concrete, and the comparison of domain-specific BERT models against BlueBERT is directionally believable. The authors even note, in Section 4.2.1, that models may have indirectly learned annotation patterns aligned with the Gold Standard because annotators were trained on it. That is a good sign of honesty.

The soft spot is evaluation independence. The headline numbers — NER F1 82.19 on Gold Standard, RE F1 66.10 — are measured on a single article produced by the same expert group that trained the annotators. That set is not an independent test set. The Test-set numbers, 61.66 NER and 51.27 RE, come from an 80/10/10 document-level split of 554 documents derived from only 76 articles. With roughly 7.3 documents per article, a random split is very likely to put segments of the same article in train and test, so even the Test-set numbers may overstate generalization. There are also no confidence intervals, no seeds, and no article-level split. None of this invalidates the corpus, but it means the reported F1 values should not be taken at face value.

The right fix is straightforward: split at the article level, run multiple seeds, and either drop the Gold Standard as a headline result or reframe it as a diagnostic set. Release the data with explicit split metadata so others can reproduce the numbers.

My recommendation: send this to peer review, but flag the evaluation protocol as the main issue. The corpus deserves to be in the literature, and the flaws are fixable. A good referee would push for re-estimation on a genuinely unseen test set.

Referee Report

4 major / 6 minor

Summary. The paper presents a corpus and an information-extraction pipeline for oxygen reduction reaction (ORR) catalyst literature in fuel cells. The authors collect 76 articles from the Royal Society of Chemistry, define 12 entity types and 2 relation types, and produce 554 documents with 16,301 entities and 13,899 relations via a two-stage annotation process using CDE parsers and Brat. They fine-tune seven BERT-based models inside the DyGIE++ framework and report NER F1 up to 82.19% and RE F1 up to 66.10% on a Gold Standard set, with lower scores (NER 61.66%, RE 51.27%) on a Test set. The paper also describes a web-based platform for collection, annotation, and extraction, and states that extracted data are publicly available. The central claim is that domain-specific BERT models outperform general or biomedical models such as BlueBERT for this task.

Significance. The corpus and pipeline, if evaluated rigorously, would be a useful contribution to materials informatics, providing a structured dataset for ORR catalysis and a demonstration of fine-tuned transformer models for domain-specific NER and RE. Strengths include the construction of a substantial manually annotated corpus, the public release of extracted data, a systematic comparison of multiple BERT variants, and the integration of a web-based tool. However, the reported F1 values are not credible estimates of generalization to unseen literature because the evaluation protocol does not ensure independence between training and test data, and the authors themselves acknowledge the potential advantage on the Gold Standard in Section 4.2.1. The qualitative ranking of domain-specific models over BlueBERT may survive a corrected evaluation, but the quantitative headline numbers require re-estimation with a properly held-out set.

major comments (4)

[Sections 3.3.1 and 4.1] The dataset of 554 documents was created by segmenting only 76 full-text articles, and the 80/10/10 train/validation/test split was applied at the document level, not the article level. Because multiple documents can originate from the same article, a random document-level split will almost certainly place segments of the same article in both training and test sets. This allows the models to exploit article-specific vocabulary and formatting, so the Test-set F1 scores in Table 5 (NER 61.66%, RE 51.27%) are not a valid estimate of performance on unseen articles. The authors should re-split at the article level, ensuring that no article contributes to more than one split, and report the resulting F1 scores.
[Sections 3.2.3 and 4.2.1] The Gold Standard was created by three experts in the same group, and the annotators whose labels form the training data were trained against this Gold Standard. Section 4.2.1 concedes that the models "may have indirectly learned annotation patterns aligned with the Gold Standard as a baseline, potentially giving them an advantage on this set." Consequently, the Gold Standard evaluation (which yields the headline NER F1 of 82.19% and RE F1 of 66.10%) is not independent of the training process. The authors must state explicitly whether the Gold Standard article(s) are part of the 76 annotated articles, and if so, exclude them from training and re-evaluate; if not, they must justify why the Gold Standard is not contaminated by the annotation-guideline alignment.
[Section 4.3] The comparison between human annotators and models is performed on a single article ("the same article used to create the Gold Standard dataset"). A single-document evaluation cannot support the general statement that "models achieved comparable performance to human annotators for fuel cell literature extraction," because both human and model performance are known to vary across documents and topics. The authors should either evaluate on a larger, representative sample of articles or restrict the claim to the specific article used in the comparison.
[Section 4.2 and Table 5] No confidence intervals, standard deviations, or multiple-seed results are reported. Given the small differences among some models (for example, 61.66% vs. 61.36% NER on the Test set), the reported F1 values alone do not establish a statistically reliable ranking. I request that the authors run multiple random seeds or at least multiple train/validation/test splits and report mean and standard deviation for each model and split.

minor comments (6)

[Abstract and Section 4.2.2] BlueBERT is a biomedical model, not a "general scientific model" as stated in the abstract and Section 4.2.2; SciBERT is the general scientific model in the comparison. Please correct this terminology.
[Section 4.1] The text "MatScicBERT-Finetuned-SQuAD-PyTorch" contains a typo ("Scic" should likely be "Sci"). Also, the distinctions among MatSciBERT-1, MatSciBERT-2, and MatSciBERT-3 would be clearer if presented in a dedicated table.
[Section 3.1.2] The procedure for ranking the 1,259 articles by "highest number of mentions of mass activity and ORR activity" is not described. Please specify whether the count is computed with a parser or simple string matching, and how ties are resolved.
[Section 4.1] The size of the Gold Standard set (number of documents, sentences, entities, and relations) is not reported, which makes it difficult to interpret the corresponding F1 scores.
[Table 1] The "Value" entity includes both numerical values with units and trend words such as "decrease" (as in the example "1246.8 m2 g-1, 1600 rpm, decrease"). Mixing numeric quantities with qualitative degree-of-change words may make the NER task harder to interpret; please clarify the annotation guideline for this entity.
[Appendix A and Data availability] The data availability statement says the data is on Mendeley Data, while Appendix A gives a URL to a CSV; please unify these descriptions and provide a persistent DOI or link.

Circularity Check

2 steps flagged · score 6.0 of 10

Headline F1 values are not clean unseen-data estimates: the Gold Standard used for benchmarking also trained the annotators whose labels train the models, and document-level splitting can leak the same article into train and test.

fitted input called prediction [Section 3.2.3 and Section 4.2.1]
"At the beginning of the annotation process, three experts in our group created a benchmark dataset, referred to as the Gold Standard. The accuracy and consistency of the annotators were evaluated by comparing their annotations against this Gold Standard. ... Additionally, since the training data was annotated by non-experts who were trained using the Gold Standard annotations, the models may have indirectly learned annotation patterns aligned with the Gold Standard as a baseline, potentially giving them an advantage on this set."

The Gold Standard is created first, used to train and evaluate the annotators whose labels form the training data, and then reused as the evaluation set for the headline NER/RE F1 values. The paper itself concedes that models 'may have indirectly learned annotation patterns aligned with the Gold Standard.' Thus the reported 82.19% NER and 66.10% RE are not independent measurements on unseen data; the test benchmark is the same benchmark that shaped the training labels, so the prediction is partly determined by the test set through the annotation pipeline.
other [Section 3.3.1 and Section 4.1]
"3. Document Splitting : Dividing documents into smaller segments to prevent CUDA out-of-memory errors ... We annotated 76 articles, resulting in a total of 554 documents in the fuel cell dataset (Table 4) after converting them into a DyGIE++ compatible format. The dataset was split into training (80%), validation (10%), and test (10%) sets."

Splitting is performed after 76 articles have been converted into 554 documents, so an 80/10/10 random document-level split can place segments of the same published article in both training and test. The Test-set F1 values (e.g., 61.66% NER and 51.27% RE) are therefore not a clean estimate of generalization to unseen articles, undercutting the Section 4.2.1 claim that the Test and Gold Standard sets 'reflect real-world performance.'

full rationale

The paper's central empirical claims are the NER/RE F1 scores and the ranking of domain-specific BERT models. The Gold Standard evaluation is circular in a partial but concrete sense: the experts created the Gold Standard, annotators were trained against it, the annotators' labels became the training data, and the same Gold Standard was then used as the benchmark for the headline F1 values. This is not full circularity because the model outputs do not generate the labels and the corpus itself is a genuine annotation effort, but it means the Gold Standard scores are inflated by a training/test dependency. The document-level split adds a second leakage pathway: because 76 articles yield 554 documents before the split, the same article can appear in both training and test, so even the lower Test-set numbers may overstate generalization. The self-citations to the authors' prior corpus work are present but not load-bearing for the ORR extraction claims, so they do not raise the score. Overall, the evaluation protocol undermines the 'unseen data' interpretation of the reported numbers, giving a partial circularity score of 6.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper is an empirical NLP benchmark; its central claims rest on annotation schema choices, evaluation protocol, and unreported training details rather than on fitted physical parameters. The listed free parameters are preprocessing and selection choices, and the axioms are domain assumptions about schema adequacy and test-set independence.

free parameters (3)

Fine-tuning hyperparameters = not reported
Section 3.3.2 names the models but gives no learning rate, batch size, epoch count, or random seed; the reported F1 values depend on these choices.
Document split size = not reported
Section 3.3.1 splits documents into smaller segments to avoid CUDA out-of-memory errors; the segment size is not given and determines whether articles cross the train/test boundary.
Article selection cutoff = 76
Section 3.1.2 selects 76 of 1259 articles by ranking mentions of mass activity and ORR activity; the cutoff is arbitrary and shapes corpus composition.

assumptions (4)

domain assumption The 12 entity types and 2 relation types are sufficient to capture ORR catalyst information.
Tables 1 and 2 define the schema; no coverage analysis or comparison with alternative schemas is provided.
ad hoc to paper The Gold Standard annotations are correct ground truth.
Section 3.2.3 states three in-group experts created the Gold Standard; no adjudication process or inter-annotator agreement is reported.
domain assumption Document-level splitting does not leak article-level content into training and test.
Section 3.3.1 splits documents and Section 4.1 applies an 80/10/10 split, but the split unit is not stated to be article; if segments from one article fall in both train and test, scores are inflated.
domain assumption The 76 selected RSC articles are representative of ORR catalyst literature.
Section 3.1.2 restricts retrieval to RSC and ranks articles by keyword mentions, introducing selection bias.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Extracting ORR Catalyst Information for Fuel Cell from Scientific Literature." pith.science (2026). https://pith.science/paper/5ML6XCPH

@misc{pith2026250707499,
  author       = {Pith},
  title        = {Pith review of: Extracting ORR Catalyst Information for Fuel Cell from Scientific Literature},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5ML6XCPH}},
  note         = {Machine review of arXiv:2507.07499}
}

read the original abstract

The oxygen reduction reaction (ORR) catalyst plays a critical role in enhancing fuel cell efficiency, making it a key focus in material science research. However, extracting structured information about ORR catalysts from vast scientific literature remains a significant challenge due to the complexity and diversity of textual data. In this study, we propose a named entity recognition (NER) and relation extraction (RE) approach using DyGIE++ with multiple pre-trained BERT variants, including MatSciBERT and PubMedBERT, to extract ORR catalyst-related information from the scientific literature, which is compiled into a fuel cell corpus for materials informatics (FC-CoMIcs). A comprehensive dataset was constructed manually by identifying 12 critical entities and two relationship types between pairs of the entities. Our methodology involves data annotation, integration, and fine-tuning of transformer-based models to enhance information extraction accuracy. We assess the impact of different BERT variants on extraction performance and investigate the effects of annotation consistency. Experimental evaluations demonstrate that the fine-tuned PubMedBERT model achieves the highest NER F1-score of 82.19% and the MatSciBERT model attains the best RE F1-score of 66.10%. Furthermore, the comparison with human annotators highlights the reliability of fine-tuned models for ORR catalyst extraction, demonstrating their potential for scalable and automated literature analysis. The results indicate that domain-specific BERT models outperform general scientific models like BlueBERT for ORR catalyst extraction.

Figures

Figures reproduced from arXiv: 2507.07499 by the authors.

**Figure 2.** Selected articles for data annotation. 3.2.1. Adopted Tools & Techniques To perform the data annotation for this study, we utilized the following tools and techniques: • CDE’s Default Parser: CDE includes a default parser capable of recognizing chemical compositions, compounds, and material elements. Additionally, CDE allows for integrating custom parsers to fit to the present study as described below. • Custom Pars… view at source ↗

**Figure 3.** Entity-Relation map. 3.2.3. Data Annotation Process At the beginning of the annotation process, three experts in our group created a benchmark dataset, referred to as the Gold Standard. The accuracy and consistency of the annotators were evaluated by comparing their annotations against this Gold Standard. The annotation process consists of two main steps: 1. Generate Default Annotations: Initial annotations in Brat… view at source ↗

Figures from the paper (9 more)

**Figure 4.** Figure 4: An example of default annotations upon a text data taken from [20]. [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]

**Figure 5.** Figure 5: An example of refine default annotations on Brat by annotator upon a text data [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]

**Figure 6.** Figure 6: Fuel cell data integration process. 3.3.2. Data Modeling Process The data modeling process focuses on training and fine-tuning machine learning models to extract named entities and relationships effectively from the annotated dataset. 13 [PITH_FULL_IMAGE:figures/full_…

**Figure 7.** Figure 7: Model selection & input data for extraction. The text data is taken from [20] [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]

**Figure 8.** Figure 8: An example of the extracted data in Brat visualization for the text data taken [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]

**Figure 9.** Figure 9: Extracted data in graph visualization. 4.1. Experiment Setup We annotated 76 articles, resulting in a total of 554 documents in the fuel cell dataset ( [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]

**Figure 10.** Figure 10: Test and Gold Standard F1-scores. • These models demonstrated the highest performance across all evaluation metrics. • The remaining models showed relatively lower effectiveness in fuel cell related NLP tasks. The ranking highlights that domain-specific BERT variants…

**Figure 11.** Figure 11: NER performance comparison [PITH_FULL_IMAGE:figures/full_fig_p024_11.png]

**Figure 12.** Figure 12: RE performance comparison affect the generalization of the models to new and unseen data. Furthermore, the model struggles with complex cross-sentence relations, which can impact the accuracy of relation extraction. Future work will aim to expand the dataset by incorp…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 29 canonical work pages

[1]

Yun Wang, Ken S. Chen, Jeffrey Mishler, Sung Chan Cho, Xavier Cor- dobes Adroher, A review of polymer electrolyte membrane fuel cells: Technology, applications, and needs on fundamental research , Applied Energy, Vol.88:981-1007, Elsevier, 2011

work page 2011
[2]

46:1848-1857, American Chemical Society, 2013

Jianbo Wu and Hong Yang, Platinum-based oxygen reduction electro- catalysts, Accounts of Chemical Research, Vol. 46:1848-1857, American Chemical Society, 2013. 24

work page 2013
[3]

Jiayao Cui, Qingjun Chen, Xiaojin Li, Suojiang Zhang, Recent advances in non-precious metal electrocatalysts for oxygen reduction in acidic me- dia and PEMFCs: an activity, stability and mechanism study , Green Chemistry, Vol.23:6898-6925, The Royal Society of Chemistry, 2021

work page 2021
[4]

Olivetti, Gerbrand Ceder, Opportunities and challenges of text mining in materials research, iScience, Vol

Olga Kononova, Tanjin He, Haoyan Huo, Amalie Trewartha, Elsa A. Olivetti, Gerbrand Ceder, Opportunities and challenges of text mining in materials research, iScience, Vol. 24:102155, 2021

work page 2021
[5]

544:101768, Elsevier, 2022

Kyosuke Yamaguchi, Ryoji Asahi, Yutaka Sasaki, Superconductivity in- formation extraction from the literature: a new corpus and its evalu- ations, Advanced Engineering Informatics, Vol. 544:101768, Elsevier, 2022

work page 2022
[6]

Rui Zhang, Jiawang Zhang, Qiaochuan Chen, Bing Wang, Yi Liu, Quan Qian, Deng Pan, Jinhua Xia, Yinggang Wang, Yuexing Han, A literature-mining method of integrating text and table extraction for materials science publications , Computational Materials Science, Vol.230:112441, Elsevier, 2023

work page 2023
[7]

Koch, Jos´ e A

Mara Schilling-Wilhelmi, Marti˜ no R ´ ıos-Garc ´ ıa, Sherjeel Shabih, Mar ´ ıa Victoria Gil, Santiago Miret, Christoph T. Koch, Jos´ e A. M´ arquez, Kevin Maik Jablonka, From text to insight: large language models for chemical data extraction, Chemical Society Reviews, Vol. 54:1125-1150, The Royal Society of Chemistry, 2025

work page 2025
[8]

15:10600-10611, The Royal Society of Chemistry, 2024

Wei Zhang, Qinggong Wang, Xiangtai Kong, Jiacheng Xiong, Shengkun Ni, Duanhua Cao, Buying Niu, Mingan Chen, Yameng Li, Runze Zhang, Yitian Wang, Lehan Zhang, Xutong Li, Zhaoping Xiong, Qian Shi, Ziming Huang, Zunyun Fu, Mingyue Zheng, Fine-tuning large language models for chemical text mining, Chemical Science, Vol. 15:10600-10611, The Royal Society of Ch...

work page 2024

Show all 29 references

[9]

Ankan Mullick, Akash Ghosh, G. Sai Chaitanya, Samir Ghui, Tapas Nayak, Seung-Cheol Lee, Satadeep Bhattacharjee, Pawan Goyala, MatSciRE: Leveraging pointer networks to automate entity and relation extraction for material science knowledge-base construction , Computa- tional Mat...

2024
[10]

Rosen, Gerbrand Ceder, Kristin A

John Dagdelen, Alexander Dunn, Sanghoon Lee, Nicholas Walker, An- drew S. Rosen, Gerbrand Ceder, Kristin A. Persson, Anubhav Jain, Structured information extraction from scientific text with large language models, Vol. 15, Nature Communications, 2024

2024
[11]

Weston, V

L. Weston, V. Tshitoyan, J. Dagdelen, O. Kononova, A. Trewartha, K. A. Persson, G. Ceder, A. Jain, Named entity recognition and normal- ization applied to large-scale information extraction from the materials science literature, Journal of Chemical Information and Modeling, Vo...

2019
[12]

618:129171, Elsevier, 2025

Nandita Goyal, Navdeep Singh: Named entity recognition and rela- tionship extraction for biomedical text, A comprehensive survey, recent advancements, and future research directions , Neurocomputing, Vol. 618:129171, Elsevier, 2025

2025
[13]

14, Nature Communications, 2023

Manu Suvarna, Alain Claude Vaucher, Sharon Mitchell, Teodoro Laino, Javier P´ erez-Ram ´ ırez,Language models and protocol standardization guidelines for accelerating synthesis planning in heterogeneous catalysis , Vol. 14, Nature Communications, 2023

2023
[14]

15:12200-12233, The Royal Society of Chemistry, 2024

Yuming Su, Xue Wang, Yuanxiang Ye, Yibo Xie, Yujing Xu, Yibin Jiang, Cheng Wang, Automation and machine learning augmented by large language models in a catalysis study , Chemical Science, Vol. 15:12200-12233, The Royal Society of Chemistry, 2024

2024
[15]

Sunoj, A transfer learning protocol for chem- ical catalysis using a recurrent neural network adapted from natural lan- guage processing, Digital Discovery, pp

Sukriti Singh, Raghavan B. Sunoj, A transfer learning protocol for chem- ical catalysis using a recurrent neural network adapted from natural lan- guage processing, Digital Discovery, pp. 303-312, The Royal Society of Chemistry, 2022

2022
[16]

28:1023-1030, Dordrecht, 2014

Williams Antony, Tkachenko Valery, The royal society of chemistry and the delivery of chemistry data repositories for the community , Journal of Computer - Aided Molecular Design, Vol. 28:1023-1030, Dordrecht, 2014

2014
[17]

Court, Taketomo Isazawa, Stephen R

Juraj Mavraˇ ci´ c, Callum J. Court, Taketomo Isazawa, Stephen R. Elliott, Jacqueline M. Cole, ChemDataExtractor 2.0: autopopulated ontologies for materials science , Journal of Chemical Information and Modeling, Vol. 61:4280-4289, American Chemical Society, 2021. 26

2021
[18]

Grinberg Miguel, Flask web development , O’Reilly Media Inc., 2018

2018
[19]

Pontus Stenetorp, Sampo Pyysalo, Goran Topi´ c, Tomoko Ohta, Sophia Ananiadou, Jun’ichi Tsujii, BRAT: a web-based tool for NLP-assisted text annotation, In Proceedings of the Demonstrations at the 13th Con- ference of the European Chapter of the Association for Computational L...

2012
[20]

16:2868-2876, The Royal Society of Chemistry, 2024

Lingling Zhang, Tong Liu, Xiaokang Liu, Sicheng Li, Xue Zhang, Qi- quan Luo, Tao Ding, Tao Yao, Wei Zhang, Highly dispersed ultrafine PtCo alloy nanoparticles on unique composite carbon supports for pro- ton exchange membrane fuel cells , Nanoscale, Vol. 16:2868-2876, The Roya...

2024
[21]

3:2206532, Taylor & Francis, 2023

Kento Mitsui, Yutaka Sasaki, Ryoji Asahi, Automatic knowledge ac- quisition from superconductivity information in literature , Science and Technology of Advanced Materials: Methods, Vol. 3:2206532, Taylor & Francis, 2023

2023
[22]

3036–3046, As- sociation for Computational Linguistics, 2019

Yi Luan, Dave Wadden, Luheng He, Amy Shah, Mari Ostendorf, Han- naneh Hajishirzi, A general framework for information extraction using dynamic span graphs, Proceedings of NAACL-HLT, pp. 3036–3046, As- sociation for Computational Linguistics, 2019

2019
[23]

3615–3620, Association for Computational Linguistics, 2019

Iz Beltagy, Kyle Lo, Arman Cohan: SciBERT, A pretrained language model for scientific text , Proceedings of the 2019 Conference on Empir- ical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pp. 3...

2019
[24]

Tanishq Gupta, Mohd Zaki, N. M. Anoop Krishnan, Mausam: MatSciB- ERT, A materials domain language model for text mining and informa- tion extraction, npj Computational Materials, Vol. 8, 2022

2022
[25]

3:1-23, Association for Computing Machinery (ACM), 2021

Yu Gu, Robert Tinn, Hao Cheng, Michael Lucas, Naoto Usuyama, Xi- aodong Liu, Tristan Naumann, Jianfeng Gao, Hoifung Poon, Domain- specific language model pretraining for biomedical natural language pro- cessing, ACM Transactions on Computing for Healthcare, Vol. 3:1-23, Associ...

2021
[26]

58-65, Association for Computational Linguistics, 2019

Yifan Peng, Shankai Yan, Zhiyong Lu, Transfer learning in biomedi- cal natural language processing: an evaluation of BERT and ELMo on ten benchmarking datasets , Proceedings of the 18th BioNLP Workshop and Shared Task, pp. 58-65, Association for Computational Linguistics, 2019

2019
[27]

MatSciBERT, https://huggingface.co/m3rg-iitd/matscibert, Ac- cessed: March 27, 2025

2025
[28]

MatSciBERT-CNER, https://huggingface.co/nlp-magnets/ matscibert-cner, Accessed: March 27, 2025

2025
[29]

MatSciBERT-Finetuned-SQuAD-PyTorch, https://huggingface.co/ rachen/matscibert-finetuned-squad-pytorch , Accessed: March 27, 2025. 28

2025

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Yun Wang, Ken S. Chen, Jeffrey Mishler, Sung Chan Cho, Xavier Cor- dobes Adroher, A review of polymer electrolyte membrane fuel cells: Technology, applications, and needs on fundamental research , Applied Energy, Vol.88:981-1007, Elsevier, 2011

work page 2011

[2] [2]

46:1848-1857, American Chemical Society, 2013

Jianbo Wu and Hong Yang, Platinum-based oxygen reduction electro- catalysts, Accounts of Chemical Research, Vol. 46:1848-1857, American Chemical Society, 2013. 24

work page 2013

[3] [3]

Jiayao Cui, Qingjun Chen, Xiaojin Li, Suojiang Zhang, Recent advances in non-precious metal electrocatalysts for oxygen reduction in acidic me- dia and PEMFCs: an activity, stability and mechanism study , Green Chemistry, Vol.23:6898-6925, The Royal Society of Chemistry, 2021

work page 2021

[4] [4]

Olivetti, Gerbrand Ceder, Opportunities and challenges of text mining in materials research, iScience, Vol

Olga Kononova, Tanjin He, Haoyan Huo, Amalie Trewartha, Elsa A. Olivetti, Gerbrand Ceder, Opportunities and challenges of text mining in materials research, iScience, Vol. 24:102155, 2021

work page 2021

[5] [5]

544:101768, Elsevier, 2022

Kyosuke Yamaguchi, Ryoji Asahi, Yutaka Sasaki, Superconductivity in- formation extraction from the literature: a new corpus and its evalu- ations, Advanced Engineering Informatics, Vol. 544:101768, Elsevier, 2022

work page 2022

[6] [6]

Rui Zhang, Jiawang Zhang, Qiaochuan Chen, Bing Wang, Yi Liu, Quan Qian, Deng Pan, Jinhua Xia, Yinggang Wang, Yuexing Han, A literature-mining method of integrating text and table extraction for materials science publications , Computational Materials Science, Vol.230:112441, Elsevier, 2023

work page 2023

[7] [7]

Koch, Jos´ e A

Mara Schilling-Wilhelmi, Marti˜ no R ´ ıos-Garc ´ ıa, Sherjeel Shabih, Mar ´ ıa Victoria Gil, Santiago Miret, Christoph T. Koch, Jos´ e A. M´ arquez, Kevin Maik Jablonka, From text to insight: large language models for chemical data extraction, Chemical Society Reviews, Vol. 54:1125-1150, The Royal Society of Chemistry, 2025

work page 2025

[8] [8]

15:10600-10611, The Royal Society of Chemistry, 2024

Wei Zhang, Qinggong Wang, Xiangtai Kong, Jiacheng Xiong, Shengkun Ni, Duanhua Cao, Buying Niu, Mingan Chen, Yameng Li, Runze Zhang, Yitian Wang, Lehan Zhang, Xutong Li, Zhaoping Xiong, Qian Shi, Ziming Huang, Zunyun Fu, Mingyue Zheng, Fine-tuning large language models for chemical text mining, Chemical Science, Vol. 15:10600-10611, The Royal Society of Ch...

work page 2024