Dataset Viewer
Auto-converted to Parquet Duplicate
record_id
stringlengths
24
36
publication_number
stringlengths
10
16
language
stringclasses
5 values
section
stringclasses
2 values
text
stringlengths
703
858k
ground_truth
listlengths
0
4.56k
chemner-CN108452553B-description
CN108452553B
CN
description
技术领域<b class="d_n">[0001]</b>本发明涉及气凝胶生产应用领域,特别是一种气凝胶用醇水回收装置。背景技术<b class="d_n">[0002]</b>气凝胶是在保持凝胶骨架结构完整的情况下,将凝胶内溶剂干燥后的产物,其是一种具有纳米多孔结构的非晶态低密度材料。二氧化硅气凝胶又被称作“蓝烟”、“固体烟”,是目前已知的最轻的固体材料,也是迄今为止保温性能最好的材料,被誉为“改变世界的神奇材料”。<b class="d_n">[0003]</b>现下,硅气凝胶的制备多是使用单一法制备,在进行超临界干燥前,需使用无水乙醇对凝胶中的溶剂进行置换,将其中的水置换出来,以降低在超临界干燥产生的毛线管力;一般置换后,都会...
[ { "start": 144, "end": 148, "text": "二氧化硅", "label": "ChemicalCompound", "smiles": "O=[Si]=O", "inchikey": "VYPSYNLAJGMNEJ-UHFFFAOYSA-N", "mention_form": "SYSTEMATIC_NAME" } ]
chemner-CN108452553B-claim
CN108452553B
CN
claim
<div class="indep-clm" num="1"><seg-con>1.一种气凝胶用醇水回收装置,包括补热器、乙醇脱水塔、分子筛、冷凝器、缓存罐、分离罐、水中转箱和无水乙醇中转箱,</seg-con><seg-trans>其特征在于,</seg-trans><seg-refi>所述补热器与乙醇脱水塔的气相进口相连,所述乙醇脱水塔的气相出口与分子筛相连,所述分子筛包括第一出口和第二出口,所述第一出口连接冷凝器,所述第二出口连接补热器;</seg-refi></div><div class="indep-clm" num="1"><seg-refi>所述冷凝器一侧连接真空泵,其出口连接缓存罐;所述分离罐内设有隔板和内筒,所述...
[]
chemner-CN117229131A-description
CN117229131A
CN
description
技术领域<b class="d_n">[0001]</b>本发明涉及农药领域,具体涉及一种1-苯基-3-(3-三氟甲基苯基)-2-丙酮的制备方法。背景技术<b class="d_n">[0002]</b>1-苯基-3-(3-三氟甲基苯基)-2-丙酮是合成氟啶草酮的一个重要中间体,氟啶草酮也称为氟啶酮(Fluridone),属于吡咯烷酮类除草剂,可用于防除大多数水中或水面杂草,也可作为芽前选择性除草剂,用于防除棉花一年生窄叶、阔叶杂草及某些多年杂草,还适用于某些对草甘膦产生抗性的杂草,如长芒苋等。<b class="d_n">[0003]</b>目前,1-苯基-3-(3-三氟甲基苯基)-2-丙酮的制备通常通过间三氟甲基苯乙腈和苯乙酸乙...
[ { "start": 3353, "end": 3355, "text": "硫酸", "label": "ChemicalCompound", "smiles": "OS(O)(=O)=O", "inchikey": "QAOWNCQODCNURD-UHFFFAOYSA-N", "mention_form": "SYSTEMATIC_NAME" }, { "start": 3359, "end": 3361, "text": "醋酸", "label": "ChemicalCompound", "smiles": "CC...
chemner-CN117229131A-claim
CN117229131A
CN
claim
<div class="indep-clm" num="1"><seg-con>1.一种1-苯基-3-(3-三氟甲基苯基)-2-丙酮的制备方法,</seg-con><seg-trans>其特征在于,</seg-trans><seg-comp>该方法包括以下步骤,</seg-comp></div><div class="indep-clm" num="1"><seg-comp>1)在甲醇钠存在下,使间三氟甲基苯乙腈与苯乙酸甲酯进行Claisen缩合反应,得到缩合产物;</seg-comp></div><div class="indep-clm" num="1"><seg-comp>2)将步骤1)得到的缩合产物进行酸水解。</seg-c...
[ { "start": 44, "end": 66, "text": "1-苯基-3-(3-三氟甲基苯基)-2-丙酮", "label": "ChemicalCompound", "smiles": "FC(F)(F)C1=CC=CC(CC(=O)CC2=CC=CC=C2)=C1", "inchikey": "MHBVFQPMQKODRK-UHFFFAOYSA-N", "mention_form": "SYSTEMATIC_NAME" }, { "start": 200, "end": 208, "text": "间三氟甲基苯乙腈", ...
chemner-US20110151089A1-description
US20110151089A1
EN
description
<b class="d_n">[0001]</b>This application is a Continuation of, and claims priority under 35 U.S.C. §120 to, International Application PCT/JP2009/062191, filed Jul. 3, 2009, and claims priority therethrough under 35 U.S.C. §119 to Japanese patent application No. 2008-175709, filed on Jul. 4, 2008, the entireties of whi...
[ { "start": 7304, "end": 7323, "text": "magnesium glutamate", "label": "ChemicalCompound", "smiles": "[Mg++].N[C@@H](CCC([O-])=O)C([O-])=O", "inchikey": "WNPYNOGQURLHSF-DFWYDOINSA-L", "mention_form": "SYSTEMATIC_NAME" }, { "start": 7948, "end": 7962, "text": "glutamic acid...
chemner-US20110151089A1-claim
US20110151089A1
EN
claim
<div class="indep-clm" num="1"><seg-con>1. A composition </seg-con><seg-trans>comprising </seg-trans><seg-refi>magnesium glutamate which can enhance the saltiness of food or drink.</seg-refi></div><div class="indep-clm" num="2"><seg-con>2. A method of enhancing the saltiness of a food or a drink </seg-con><seg-trans>co...
[ { "start": 111, "end": 130, "text": "magnesium glutamate", "label": "ChemicalCompound", "smiles": "[Mg++].N[C@@H](CCC([O-])=O)C([O-])=O", "inchikey": "WNPYNOGQURLHSF-DFWYDOINSA-L", "mention_form": "SYSTEMATIC_NAME" }, { "start": 394, "end": 414, "text": " magnesium glutam...
chemner-US10065976B2-description
US10065976B2
EN
description
"<b class=\"d_n\">[0001]</b>This application claims the benefit of U.S. Provisional Application No. (...TRUNCATED)
[{"start":73515,"end":73532,"text":" CDCl<sub>3</sub>","label":"ChemicalCompound","smiles":"[2H]C(Cl(...TRUNCATED)
chemner-US10065976B2-claim
US10065976B2
EN
claim
"<div class=\"indep-clm\" num=\"1\"><seg-con>1. A method of forming a pyrrole or pyrazole compound t(...TRUNCATED)
[{"start":65,"end":72,"text":"pyrrole","label":"ChemicalCompound","smiles":"N1C=CC=C1","inchikey":"K(...TRUNCATED)
chemner-KR1020220102156A-description
KR1020220102156A
KR
description
"<b class=\"d_n\">[0001]</b><b>관련 출원에 대한 상호 참조</b>\n<b class=\"d_n\">[0002]</b(...TRUNCATED)
[{"start":451377,"end":451422,"text":"삼차-부틸 (2S,6R)-2-(하이드록시메틸)-6-메틸-모(...TRUNCATED)
chemner-KR1020220102156A-claim
KR1020220102156A
KR
claim
"<div class=\"indep-clm\" num=\"1\">청구항 1</div>\n<div class=\"indep-clm\" num=\"1\">하기 화(...TRUNCATED)
[{"start":672,"end":674,"text":"Cl","label":"ChemicalCompound","smiles":"","inchikey":"","mention_fo(...TRUNCATED)
End of preview. Expand in Data Studio

Chemical NER Bench

English | 简体中文

PatentLSBench

Chemical NER Bench provides complete patent claims and descriptions for chemical entity recognition and structure normalization, with entity-level reference answers and molecular structure identifiers.

Task description · Evaluation guide · Schema

Dataset overview

V1.0 contains:

Resource Count
Patent publication numbers 11
Complete section inputs 22
Claims / description inputs 11 / 11
Entity reference mentions 6,334
Inputs with an empty entity reference list 3

The dataset covers Chinese, English, Japanese and Korean, including Traditional Chinese text.

Language Code Sections
Chinese CN 8
Traditional Chinese CNT 2
English EN 6
Japanese JP 4
Korean KR 2

Input and reference answers

Each record is one complete claim or description section, evaluated as an independent input. Model inputs contain five fields: record_id, publication_number, language, section and text. ground_truth contains entity-level reference answers for comparison with model predictions.

Fields

Record field Meaning
record_id Unique identifier for a section record.
publication_number Publication number of the patent containing this section.
language Section language code: CN = Chinese, CNT = Traditional Chinese, EN = English, JP = Japanese, KR = Korean.
section Section type: claim = complete claims section; description = complete description section.
text HTML text of a complete claims or description section.
ground_truth Entity-level evaluation reference answers, with one object per entity mention; [] when there are no reference entities.
Entity field Meaning
start Inclusive, zero-based start offset of the entity in this record's text, measured in Unicode code points including HTML markup characters.
end Exclusive end offset of the entity in this record's text, forming a half-open interval with start.
text Entity name text in the section.
label Chemical entity class code. ChemicalCompound and compound both denote chemical entities and do not distinguish chemical object subtypes.
smiles SMILES representation of the entity's reference structure; an empty string indicates no available SMILES reference.
inchikey InChIKey identifier of the entity's reference structure; an empty string indicates no available InChIKey reference.
mention_form How the entity name is expressed in the text; values are defined below. null indicates no name-form category information.

ChemicalCompound and compound both denote chemical entities rather than different chemical object subtypes. Record text contains the complete section HTML. Offsets count all HTML characters in Unicode code points using zero-based, half-open [start, end) intervals.

ground_truth = [] indicates a section with no reference entities. Empty structure strings indicate no reference for the corresponding identifier; mention_form = null indicates no name-form category information.

Name forms

Value Name Meaning
SYSTEMATIC_NAME Full name Full or explicit structural name, including IUPAC names and short names with a definite chemical identity.
TRIVIAL_COMMON_NAME Common name Conventional chemical name, generic drug name, or trade name identifying a single ingredient.
UNAMBIGUOUS_ABBREVIATION Abbreviation Name abbreviation with an unambiguous chemical identity in context, such as DMSO or DCM. Development codes, identifiers, and amino-acid sequence symbols are excluded.
UNAMBIGUOUS_FORMULA Condensed structural formula Unambiguous condensed structural formula or chemical element symbol, such as CH3COOH or Cu. Molecular formulas that do not identify a definite structure are excluded.
null No category information No name-form category information.

The schema defines field types, allowed values and required fields.

Evaluation

Reference coverage, representation differences and empty-reference records are described in the reference notes. The two structure fields are not interchangeable for every entity.

Entity reference answers support comparison of names, chemical entity classes and boundaries. Structure normalization uses eligible nonempty references in the selected structure field. Report precision, recall and F1 separately for entity recognition and structure normalization.

Reports specify matching units and conditions, deduplication, structure-reference eligibility and coverage, aggregation and zero-denominator conventions. Present claims and description results separately. Mentions, names and structures are distinct counting units; an entity without a structure reference remains available for applicable entity-recognition evaluation.

See the evaluation guide for reporting requirements.

Configuration and use

Hugging Face provides the default configuration with 22 records in the test split.

from datasets import load_dataset

records = load_dataset("PatSnap/chemical-ner-bench", split="test", revision="V1.0")
input_fields = [
    "record_id", "publication_number", "language", "section", "text"
]
model_inputs = records.select_columns(input_fields)
model_input = model_inputs[0]

Use ground_truth separately for evaluation, excluding it from model requests.

Path Contents
data/test-00000-of-00001.parquet Native HF data view
raw/records.jsonl Complete section records and entity references
raw/schema.json Field schema
raw/load_records.py Standard-library reader and five-field input projection
EVALUATION.md / EVALUATION.zh.md English / Chinese evaluation guide
REFERENCE_NOTES.md / REFERENCE_NOTES.zh.md Reference coverage and representation limits
validate_dataset.py Record, offset and optional structure diagnostics

Read the JSONL with the standard-library reader:

python raw/load_records.py raw/records.jsonl

Validate the supplied records and offsets (standard library only):

python validate_dataset.py

Citation

@dataset{patsnap_chemical_ner_bench_2026,
  title   = {Chemical NER Bench},
  author  = {PatSnap},
  year    = {2026},
  version = {1.0},
  url     = {https://huggingface.co/datasets/PatSnap/chemical-ner-bench}
}

License and Disclaimer

The dataset, annotations and accompanying assets use CC BY-NC 4.0. Reader code and project documentation use the MIT License; third-party materials retain their applicable rights and separate terms.

The dataset and related outputs may contain errors or omissions and do not constitute legal, patent, medical or regulatory advice or professional certification. Review and validate them before relying on them; see the full disclaimer.

Downloads last month
36