Datasets:
corpus_id int64 3.52k 274M | doi stringlengths 12 55 | url stringlengths 78 78 | title stringlengths 13 300 | authors listlengths 0 162 | author_ids listlengths 0 162 | venue stringlengths 0 165 | year int64 1.97k 2.02k | fields_of_study listlengths 2 8 | publication_date stringlengths 0 10 | journal_name stringlengths 0 165 | license_publisher stringclasses 482
values | license stringclasses 1
value | oa_url stringlengths 30 1.41k | text stringlengths 1 10k | field_classification listlengths 0 5 | text_type stringclasses 2
values | embedding listlengths 1.02k 1.02k | label listlengths 1 5 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
16,247,147 | 10.3390/molecules21091195 | https://www.semanticscholar.org/paper/0f73074ca2947500db3ca7e0a5d55570715d72e3 | Comparison of Protein N-Homocysteinylation in Rat Plasma under Elevated Homocysteine Using a Specific Chemical Labeling Method | [
"T. Zang",
"Ligi Paul Pottenplackel",
"D. Handy",
"J. Loscalzo",
"Shujia Dai",
"R. Deth",
"Z. Zhou",
"Jisheng Ma"
] | [
"8115255",
"89932130",
"3374868",
"2501779",
"3656891",
"2248134",
"50252016",
"46282292"
] | Molecules | 2,016 | [
"Chemistry",
"Medicine",
"Chemistry",
"Medicine"
] | 2016-09-01 | Molecules | MDPI AG | cc-by | https://www.mdpi.com/1420-3049/21/9/1195/pdf?version=1473323206 | A selective chemical method was developed to label N-homocysteinylated proteins, revealing increased N-homocysteinylation levels by 1.8-fold in vitamin B12 deficient rats. This suggests metabolic correlations with total homocysteine, highlighting additional metabolic factors affecting protein modifications. | [
"Biochemistry",
"Medicinal Chemistry"
] | summary | [
0.0013429223326966166,
-0.03110085241496563,
0.0013129721628502011,
-0.0010151705937460065,
-0.03301035612821579,
0.011139518581330776,
-0.03010522946715355,
-0.04423019662499428,
-0.030382690951228142,
0.038729917258024216,
0.04259030893445015,
-0.02494088001549244,
0.031231878325343132,
... | [
7,
10
] |
7,317,798 | 10.3390/molecules181214564 | https://www.semanticscholar.org/paper/b23c2e991ab49303834efc42bf2f5da0199c43f5 | Insights into Structure-Activity Relationships of Somatostatin Analogs Containing Mesitylalanine | [
"Pablo Martín-Gago",
"E. Aragón",
"Marc Gomez‐Caminals",
"J. Fernández‐Carneado",
"Rosario Ramón",
"P. Martin-Malpartida",
"X. Verdaguer",
"P. López-Ruiz",
"B. Colás",
"M. Cortes",
"B. Ponsati",
"M. Macias",
"A. Riera"
] | [
"1404788102",
"121554298",
"1402015991",
"1402015905",
"37411362",
"1400946938",
"4130580",
"1400873047",
"47262934",
"7029783",
"4161736",
"145380178",
"145272936"
] | Molecules | 2,013 | [
"Chemistry",
"Medicine",
"Chemistry",
"Medicine"
] | 2013-11-25 | Molecules | MDPI AG | cc-by | https://www.mdpi.com/1420-3049/18/12/14564/pdf?version=1403115326 | Somatostatin analogs containing mesitylalanine residues exhibit more restricted conformations compared to the natural hormone, with some peptides showing selective binding to somatostatin receptors and increased serum stability, while multiple incorporations of Msa reduced conformational rigidity and affinity. | [
"Biochemistry",
"Medicinal Chemistry"
] | summary | [
0.013775872997939587,
0.002446887781843543,
-0.006771123968064785,
0.0006849437486380339,
-0.05722004920244217,
0.04605281725525856,
-0.023859135806560516,
-0.04869527369737625,
-0.02665713243186474,
0.03473527729511261,
0.0322047583758831,
-0.02828594483435154,
0.012180233374238014,
0.039... | [
7,
10
] |
54,757,345 | 10.1155/2011/359161 | https://www.semanticscholar.org/paper/753b398509e166587d48969eadd4e0cdd61d8882 | The Characterization and Hydrogen Production from Water Decomposition with Methanol in a Semi-Batch Type Reactor Using In, P-TiO2s | [
"Joonwoo Kim",
"Hae-Ri Kim",
"S. Choung"
] | [
"2117059716",
"2109741453",
"14974970"
] | 2,011 | [
"Environmental Science",
"Chemistry",
"Engineering",
"Chemistry"
] | 2011-04-13 | International Journal of Photoenergy | Hindawi Limited | cc-by | https://downloads.hindawi.com/journals/ijp/2011/359161.pdf | The photocatalytic production of hydrogen from water using solar energy is potentially a clean and renewable source for hydrogen fuel. This study examines the production of hydrogen over In, P-TiO2s photocatalysts. 1 mol% In-TiO2 and P-TiO2 were produced using the solvothermal method and were treated at 500 and 800∘C t... | [
"Catalysis",
"Materials Science",
"Energy Chemistry"
] | abstract | [
0.016892084851861,
-0.03366902470588684,
0.02675066888332367,
0.019084850326180458,
-0.05073961615562439,
0.039114806801080704,
-0.018659161403775215,
-0.005465773865580559,
-0.03253308683633804,
0.01781720668077469,
0.019401732832193375,
-0.06437893956899643,
-0.0010472118156030774,
0.030... | [
0,
4,
9
] | |
237,931,822 | 10.3390/su13136997 | https://www.semanticscholar.org/paper/37ba534ce9007e1defb56750f1d1881d7b5e16c7 | Macadamia Husk Compost Improved Physical and Chemical Properties of a Sandy Loam Soil | [
"Dembe Maselesele",
"J. Ogola",
"R. Murovhi"
] | [
"2129429010",
"88580197",
"88119098"
] | Sustainability | 2,021 | [
"Agricultural and Food Sciences",
"Environmental Science",
"Chemistry"
] | 2021-06-22 | Sustainability | MDPI AG | cc-by | https://www.mdpi.com/2071-1050/13/13/6997/pdf?version=1624351221 | Macadamia husk compost improves soil fertility by increasing organic carbon, pH, and nutrient availability, effectively enhancing crop productivity in sandy loam soil in South Africa. The most significant improvements were seen with 30 t ha⁻¹ application, showing potential for sustainable soil productivity enhancement. | [
"Environmental Chemistry",
"Agricultural Science"
] | summary | [
-0.001014030771329999,
-0.038078274577856064,
0.025691045448184013,
-0.017355293035507202,
-0.04095989465713501,
0.010491074062883854,
-0.041201118379831314,
-0.010578314773738384,
-0.0355965681374073,
0.027705157175660133,
0.02654138021171093,
-0.0022572576999664307,
0.028613600879907608,
... | [
8
] |
23,738,106 | 10.1074/jbc.271.20.12009 | https://www.semanticscholar.org/paper/8f8e2604b5ce8735cb24823e199e730bdd4cae23 | High Mobility Group Proteins 14 and 17 Can Prevent the Close Packing of Nucleosomes by Increasing the Strength of Protein Contacts in the Linker DNA (*) | [
"D. Tremethick",
"Luke Hyman"
] | [
"5779380",
"2209573801"
] | Journal of Biological Chemistry | 1,996 | [
"Biology",
"Chemistry",
"Biology",
"Medicine"
] | 1996-05-17 | The Journal of Biological Chemistry | Elsevier BV | cc-by | http://www.jbc.org/article/S0021925818826527/pdf | High mobility group (HMG) proteins 14 and 17 are abundant chromatin-associated proteins found in all higher eukaryotic nuclei. This observation demonstrates that HMGs 14 and 17 must have an important and universal function with regard to the structure and function of chromatin. What this function is, including how they... | [
"Biochemistry"
] | abstract | [
-0.01816977746784687,
-0.024239256978034973,
0.05695062503218651,
0.007561890408396721,
-0.040318772196769714,
0.020588604733347893,
-0.03868180885910988,
-0.04157699644565582,
-0.04777783900499344,
0.005426398944109678,
0.028646130114793777,
-0.02612932212650776,
0.03950374200940132,
0.02... | [
7
] |
202,402,710 | 10.3390/nano9091284 | https://www.semanticscholar.org/paper/f5196ae8904f698cff47494974c6375b0d6287b2 | Effective Oxygen Reduction Reaction Performance of FeCo Alloys In Situ Anchored on Nitrogen-Doped Carbon by the Microwave-Assistant Carbon Bath Method and Subsequent Plasma Etching | [
"Mincong Liu",
"Feng Yu",
"Cunhua Ma",
"Xueyan Xue",
"Hai-hai Fu",
"Huifang Yuan",
"Shengchao Yang",
"Gang Wang",
"Xuhong Guo",
"Lili Zhang"
] | [
"2143979394",
"144053457",
"92128831",
"94936753",
"94644331",
"2020100118",
"51413224",
"48737747",
"144466282",
"2144206427"
] | Nanomaterials | 2,019 | [
"Materials Science",
"Chemistry",
"Engineering",
"Medicine",
"Materials Science"
] | 2019-09-01 | Nanomaterials | MDPI AG | cc-by | https://www.mdpi.com/2079-4991/9/9/1284/pdf?version=1567945040 | Bimetal FeCo alloy nanoparticles wrapped in N-doped graphitic carbon were synthesized for enhanced oxygen reduction reactions, achieving a half-wave potential of 0.88 V and an onset potential of 0.95 V vs. RHE, with superior selectivity and stability compared to Pt/C in alkaline media. | [
"Catalysis",
"Materials Science",
"Energy Chemistry"
] | summary | [
-0.003810238791629672,
-0.044128745794296265,
0.025614412501454353,
-0.03905940055847168,
-0.02675713784992695,
0.0019974680617451668,
-0.04705704376101494,
-0.010803365148603916,
-0.024223705753684044,
0.020346859470009804,
0.017575623467564583,
-0.016040053218603134,
0.035926997661590576,
... | [
0,
4,
9
] |
6,477,740 | 10.1371/journal.pone.0026592 | https://www.semanticscholar.org/paper/06d96cc3b926bba55f0306b7938ab67b55f0d6e9 | Efficient Generation of Fully Reprogrammed Human iPS Cells via Polycistronic Retroviral Vector and a New Cocktail of Chemical Compounds | [
"Zhonghui Zhang",
"Yongxing Gao",
"Albert Gordon",
"Zack Z Wang",
"Zhijian Qian",
"Wen‐shu Wu"
] | [
"2458262",
"4185188",
"1614042703",
"2000877747",
"5752367",
"48714300"
] | PLoS ONE | 2,011 | [
"Chemistry",
"Medicine",
"Biology",
"Medicine"
] | 2011-10-26 | PLoS ONE | Public Library of Science (PLoS) | cc-by | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0026592&type=printable | A new method using a polycistronic retroviral vector with four transcription factors and three small molecules improves reprogramming efficiency of human somatic cells into iPS cells, significantly reducing partially reprogrammed colonies and facilitating isolation of fully reprogrammed iPS cells. | [
"Biochemistry",
"Medicinal Chemistry"
] | summary | [
0.0235135555267334,
-0.030621636658906937,
-0.007172301411628723,
-0.02785191312432289,
-0.05137232691049576,
0.010295947082340717,
-0.04439013451337814,
-0.00663323700428009,
-0.03245415538549423,
0.008304123766720295,
0.031476400792598724,
-0.02020920068025589,
0.020870724692940712,
0.03... | [
7,
10
] |
232,774,380 | 10.3390/medicines8040017 | https://www.semanticscholar.org/paper/917f50a51da765d7bba5f6922b9555422cb458fa | A Unique Anti-Cancer 3-Styrylchromone Suppresses Inflammatory Response via HMGB1-RAGE Signaling | [
"H. Abe",
"Miwa Okazawa",
"T. Oyama",
"Hiroaki Yamazaki",
"A. Yoshimori",
"T. Kamiya",
"Mitsutoshi Tsukimoto",
"K. Takao",
"Y. Sugita",
"H. Sakagami",
"Takehiko Abe",
"S. Tanuma"
] | [
"47243095",
"1471696436",
"92036080",
"1879161020",
"5975150",
"1379682900",
"6716942",
"48356366",
"4607739",
"2532428",
"1994686396",
"83857228"
] | Medicines | 2,021 | [
"Medicine",
"Chemistry",
"Medicine"
] | 2021-03-24 | Medicines | MDPI AG | cc-by | https://www.mdpi.com/2305-6320/8/4/17/pdf?version=1616917309 | Abstract Background: High mobility group box 1 (HMGB1)-receptor for advanced glycation endo-products (RAGE) axis serves as a key player in linking inflammation and carcinogenesis. Recently, papaverine was revealed to suppress the HMGB1-RAGE inflammatory signaling pathway and cancer cell proliferation. Therefore, a dual... | [
"Medicinal Chemistry",
"Biochemistry"
] | abstract | [
0.004844866693019867,
-0.02274317480623722,
0.015338773839175701,
-0.020425137132406235,
-0.02234872616827488,
0.01602817513048649,
-0.031031930819153786,
-0.03458663821220398,
-0.0355740524828434,
0.019578738138079643,
0.04261467233300209,
-0.029166551306843758,
0.023242389783263206,
0.03... | [
10,
7
] |
23,031,899 | 10.1074/jbc.M805956200 | https://www.semanticscholar.org/paper/8ec836195df6125d07a4b7066b3d362deeb85b2b | Functional Protein Delivery into Neurons Using Polymeric Nanoparticles* | [
"L. Hasadsri",
"J. Kreuter",
"H. Hattori",
"T. Iwasaki",
"J. George"
] | [
"4473031",
"2471797",
"34973707",
"4777420",
"2099292"
] | Journal of Biological Chemistry | 2,009 | [
"Materials Science",
"Medicine",
"Chemistry",
"Medicine"
] | 2009-03-13 | Journal of Biological Chemistry | Elsevier BV | cc-by | http://www.jbc.org/article/S0021925820325813/pdf | "Polybutylcyanoacrylate nanoparticles efficiently deliver intact proteins into neurons via endocytos(...TRUNCATED) | [
"Biochemistry",
"Medicinal Chemistry"
] | summary | [0.009288402274250984,-0.03057197667658329,0.019184110686182976,-0.04035910591483116,-0.061083510518(...TRUNCATED) | [
7,
10
] |
236,599,859 | 10.3390/APP11104354 | https://www.semanticscholar.org/paper/f19b7df0b125267ba4d45c66397bc77a8d2526a5 | "Antifouling Membranes Based on Cellulose Acetate (CA) Blended with Poly(acrylic acid) for Heavy Met(...TRUNCATED) | ["I. Ounifi","Youssef Guesmi","C. Ursino","S. Santoro","Sélim Mahfoudhi","A. Figoli","Ezzedin Ferja(...TRUNCATED) | [
"1397267671",
"97046800",
"5770197",
"96517934",
"15684438",
"6290460",
"2121831855",
"34805920"
] | Applied Sciences | 2,021 | [
"Environmental Science",
"Chemistry",
"Materials Science",
"Chemistry"
] | 2021-05-11 | Applied Sciences | MDPI AG | cc-by | https://www.mdpi.com/2076-3417/11/10/4354/pdf?version=1620732200 | "Novel fouling-resistant ultrafiltration membranes made from cellulose acetate blended with poly(acr(...TRUNCATED) | [
"Materials Science",
"Environmental Chemistry"
] | summary | [0.007561517413705587,-0.048423413187265396,0.04381849616765976,-0.006682268343865871,-0.07161512970(...TRUNCATED) | [
4,
8
] |
Lit2Vec Subfield Classifier Dataset
This dataset contains 39,900 short chemistry texts with e5-large-v2 embeddings and multi-label subfield annotations. It is the training data of the subfield classifier of the Lit2Vec workflow (paper, code).
The texts come from 19,992 papers whose license metadata indicates CC BY. They are the same papers as in the TL;DR dataset, and the texts are identical to its abstracts and summaries. Each paper contributes two records: its abstract (19,992 records) and its GPT-4o summary (19,908 records, because 84 papers have no summary). Both records of a paper carry the same labels.
The subfield labels were generated by GPT-4o from a controlled vocabulary of 17 chemistry subfields plus "Others". They were not curated by experts. Two domain experts reviewed a random sample of 45 annotated abstracts and found them generally reasonable (see the paper).
Splits
| Split | Records |
|---|---|
| train | 31,920 |
| validation | 3,990 |
| test | 3,990 |
Fields
Each record is one line of a JSON Lines file under data/.
| Field | Type | Description |
|---|---|---|
corpus_id |
int | Semantic Scholar corpus ID. Two records share it (abstract and summary). |
doi |
string | DOI. |
url |
string | Semantic Scholar page of the paper. |
title |
string | Title. |
authors |
list of strings | Author names. Empty for 70 records. |
author_ids |
list of strings | Semantic Scholar author IDs, aligned with authors; an empty string where the ID is unknown. |
venue |
string | Venue. Empty for 7,263 records. |
year |
int | Publication year (1967 to 2024). |
fields_of_study |
list of strings | Field labels from Semantic Scholar. |
publication_date |
string | ISO date. Empty for 1,458 records. |
journal_name |
string | Journal name. Empty for 1,918 records. |
license_publisher |
string | Publisher of the open-access version. |
license |
string | License category from the screening. It is cc-by for every record. |
oa_url |
string | Link to the open-access full text. |
text |
string | The abstract or the summary, see text_type. |
text_type |
string | abstract or summary. |
field_classification |
list of strings | Subfield names as produced by GPT-4o. |
label |
list of ints | IDs of the names in field_classification that belong to the 18-class vocabulary. |
embedding |
list of 1,024 floats | e5-large-v2 embedding of text (see below). |
Embeddings
embedding is the L2-normalised output of intfloat/e5-large-v2 for the string abstract: <text> when text_type is abstract and summary: <text> when it is summary. We checked this by re-encoding records: the stored vectors and the re-computed ones have a cosine similarity of 1.00000. To classify new text with the model, use the same prefixes.
Labels
| ID | Subfield | Papers | ID | Subfield | Papers | |
|---|---|---|---|---|---|---|
| 0 | Catalysis | 997 | 9 | Energy Chemistry | 754 | |
| 1 | Organic Chemistry | 1,247 | 10 | Medicinal Chemistry | 6,326 | |
| 2 | Polymer Chemistry | 569 | 11 | Chemical Engineering | 2,019 | |
| 3 | Inorganic Chemistry | 1,053 | 12 | Supramolecular Chemistry | 207 | |
| 4 | Materials Science | 4,664 | 13 | Radiochemistry & Nuclear Chemistry | 93 | |
| 5 | Analytical Chemistry | 3,208 | 14 | Forensic & Legal Chemistry | 61 | |
| 6 | Physical Chemistry | 1,157 | 15 | Food Chemistry | 1,502 | |
| 7 | Biochemistry | 10,514 | 16 | Chemical Education | 91 | |
| 8 | Environmental Chemistry | 2,664 | 17 | Others | 125 |
The last column counts papers (out of 19,992). A paper has 1.86 labels on average, and the distribution is strongly skewed towards Biochemistry and Medicinal Chemistry.
Example
A paper with its two records (embeddings and long fields shortened):
{"corpus_id": 105403827, "text_type": "abstract",
"title": "Modeling of Chemical Reacting Transport Phenomena in a PEM Fuel Cell using Finite Volume Method",
"doi": "10.31031/PPS.2018.02.000549", "year": 2018, "license": "cc-by",
"text": "Modeling the transport phenomena in a fuel cell system is important to the development of fuel cells. ...",
"field_classification": ["Energy Chemistry", "Chemical Engineering"], "label": [9, 11],
"embedding": [-0.0153, -0.0243, 0.0022, "... 1,024 values"]}
{"corpus_id": 105403827, "text_type": "summary",
"text": "A two-dimensional transient model using finite volume method enhances PEMFC design by simulating gas flow and solving movement and energy equations.",
"field_classification": ["Energy Chemistry", "Chemical Engineering"], "label": [9, 11],
"embedding": [-0.0149, -0.0214, 0.0076, "... 1,024 values"]}
Usage
from datasets import load_dataset
ds = load_dataset("Bocklitz-Lab/lit2vec-subfield-classifier-dataset")
print(ds["train"][0]["title"])
print(ds["train"][0]["label"])
Licensing and attribution
The texts are the work of the original authors and were published under CC BY, which requires attribution. Each record carries the DOI, the publisher and a link to the open-access full text. The summaries and the labels are machine-generated, were produced with OpenAI's GPT-4o, and are released under CC BY 4.0 together with the abstracts. Check OpenAI's terms of use if you plan to use them to train other models. The license metadata was resolved from Unpaywall, OpenAlex and Crossref as described in the paper. This is a metadata-based screen and not legal clearance.
Citation
Please cite the Lit2Vec paper:
@misc{amiri2026lit2vec,
title = {Lit2Vec: A Reproducible Workflow for Building a Legally Screened Chemistry Corpus from S2ORC for Downstream Retrieval and Text Mining},
author = {Amiri, Mahmoud and Jafari, Jamile Mohammad and Mostafapour, Sara and Bocklitz, Thomas},
year = {2026},
eprint = {2604.12498},
archivePrefix = {arXiv},
doi = {10.48550/arXiv.2604.12498},
url = {https://arxiv.org/abs/2604.12498}
}
Acknowledgements
The texts come from S2ORC, provided by the Allen Institute for AI and the Semantic Scholar team. The embeddings were computed with intfloat/e5-large-v2. The work was developed in the Bocklitz Lab at the Leibniz Institute of Photonic Technology (Leibniz IPHT) and Friedrich Schiller University Jena and was funded by the BMFTR within the programme Photonics Research Germany (13N15466, 13N15710, 13N15715). Questions: mahmoud.amiri@uni-jena.de.
- Downloads last month
- 53