| { |
| "@context": { |
| "@language": "en", |
| "@vocab": "https://schema.org/", |
| "citeAs": "cr:citeAs", |
| "column": "cr:column", |
| "conformsTo": "dct:conformsTo", |
| "cr": "http://mlcommons.org/croissant/", |
| "rai": "http://mlcommons.org/croissant/RAI/", |
| "prov": "http://www.w3.org/ns/prov#", |
| "data": { |
| "@id": "cr:data", |
| "@type": "@json" |
| }, |
| "dataType": { |
| "@id": "cr:dataType", |
| "@type": "@vocab" |
| }, |
| "dct": "http://purl.org/dc/terms/", |
| "examples": { |
| "@id": "cr:examples", |
| "@type": "@json" |
| }, |
| "equivalentProperty": "cr:equivalentProperty", |
| "extract": "cr:extract", |
| "field": "cr:field", |
| "fileProperty": "cr:fileProperty", |
| "fileObject": "cr:fileObject", |
| "fileSet": "cr:fileSet", |
| "format": "cr:format", |
| "includes": "cr:includes", |
| "isLiveDataset": "cr:isLiveDataset", |
| "jsonPath": "cr:jsonPath", |
| "key": "cr:key", |
| "md5": "cr:md5", |
| "parentField": "cr:parentField", |
| "path": "cr:path", |
| "recordSet": "cr:recordSet", |
| "references": "cr:references", |
| "regex": "cr:regex", |
| "repeated": "cr:repeated", |
| "replace": "cr:replace", |
| "sc": "https://schema.org/", |
| "samplingRate": "cr:samplingRate", |
| "separator": "cr:separator", |
| "source": "cr:source", |
| "subField": "cr:subField", |
| "transform": "cr:transform" |
| }, |
| "@type": "sc:Dataset", |
| "conformsTo": [ |
| "http://mlcommons.org/croissant/1.1", |
| "http://mlcommons.org/croissant/RAI/1.0" |
| ], |
| "name": "ml-conference-paper-dataset", |
| "description": "A large-scale dataset of 47,214 research papers from five top machine learning conferences (ICLR 2024, ICLR 2025, ICLR 2026, ICML 2025, NeurIPS 2025), including full text, peer reviews, editorial decisions, GROBID-parsed metadata, and bibliographic references. Designed for research on peer review analysis, paper quality prediction, NLP for scientific literature, and meta-science.", |
| "license": "https://spdx.org/licenses/CC-BY-4.0", |
| "url": "https://huggingface.co/datasets/anoyresearcher/prism_paper_data", |
| "version": "1.0.0", |
| "datePublished": "2026-03-29", |
| "dateModified": "2026-05-04", |
| "inLanguage": ["en"], |
| "keywords": [ |
| "scientific-papers", |
| "peer-review", |
| "machine-learning", |
| "openreview", |
| "grobid", |
| "tei-xml", |
| "bibtex", |
| "natural-language-processing", |
| "meta-science" |
| ], |
| "creator": { |
| "@type": "sc:Person", |
| "name": "anoyresearcher", |
| "url": "https://huggingface.co/anoyresearcher" |
| }, |
| "publisher": { |
| "@type": "sc:Organization", |
| "name": "Anonymous AI Lab" |
| }, |
| "sameAs": "https://huggingface.co/datasets/anoyresearcher/prism_paper_data", |
| "isAccessibleForFree": true, |
| "spatialCoverage": null, |
| "temporalCoverage": "2024/2026", |
| "citeAs": "anoyresearcher. (2026). ml-conference-paper-dataset (Version 1.0.0) [Data set]. Hugging Face. https://huggingface.co/datasets/anoyresearcher/prism_paper_data", |
| "usageInfo": "Dataset metadata, OpenReview comments/configuration records, and ICML/PMLR article content are documented as CC BY 4.0 where applicable. NeurIPS paper copyright is retained by the paper authors, with NeurIPS holding publication/distribution rights; downstream users should verify per-paper source licenses before redistributing or reusing full-text paper content. This dataset is intended for research use with attribution to the original paper authors, venues, and source platforms.", |
| "copyrightNotice": "Source papers remain copyright of their respective authors unless the source venue states otherwise. Derived metadata and dataset packaging are released under CC BY 4.0.", |
| "rai:dataLimitations": [ |
| "ICML 2025 and NeurIPS 2025 data was collected from conference proceedings, not OpenReview. Only ~26.9% of ICML submissions and ~24.5% of NeurIPS submissions are included. The data is heavily biased toward accepted papers, making these venues unsuitable for acceptance prediction, reviewer bias analysis, or rejection-related studies. Only ICLR data (2024-2026) has balanced accept/reject coverage.", |
| "Approximately 20-27% of ICLR papers (1,482 in 2024, 2,793 in 2025, 5,299 in 2026) have 'Pending' as the decision with meta-review text 'TBD'. This is because the data was scraped during the review period, after initial peer reviews were posted but before meta-reviews and final decisions were announced. These papers have complete reviews and full text, but no final decision or meta-review. They should be treated as missing decision data, not as withdrawn or rejected papers.", |
| "Rating scales differ between venues: ICLR 2025/2026 use numeric ratings (1-10 or 0-10), NeurIPS uses 1-6 scale, ICLR 2024 uses text-based fields (Soundness/Presentation/Contribution: poor/fair/good/excellent) with an empty Rating field, and ICML 2025 uses Overall Recommendation. Cross-venue rating comparisons are not meaningful.", |
| "GROBID parsing errors: Some papers have incomplete or malformed GROBID output depending on PDF formatting quality. GROBID fulltext extraction is missing for a small number of papers across venues.", |
| "Some PDFs contain scanned images or non-standard layouts that reduce text extraction quality. GROBID fulltext extraction is missing for a small number of papers across venues. This dataset is strictly NOT recommended for: (1) cross-venue rating comparison, as rating scales and formats differ between venues; (2) studying NeurIPS 2025 or ICML 2025 rejection patterns, as these venues are heavily biased toward accepted papers (~95% acceptance rate in the dataset vs ~25% real acceptance rate); (3) as a representative sample of all computer science research, as only top-tier ML venues are represented; (4) any clinical, medical, or safety-critical application, as this is a research dataset only; (5) inferring reviewer identity from review patterns, as this would violate reviewer privacy." |
| ], |
| "rai:dataBiases": [ |
| "ICML 2025 and NeurIPS 2025 are heavily biased toward accepted papers (~95% acceptance rate in dataset vs ~25% real acceptance rate). Any analysis using these venues will reflect accepted paper characteristics only, not the full submission pool. These venues are suitable only for studying accepted paper characteristics, topic modeling, and bibliographic analysis.", |
| "~20-27% of ICLR papers have 'Pending' decisions because the data was scraped during the review period (after initial reviews but before final decisions). This introduces temporal bias — these papers were captured at an earlier stage of the review process compared to papers with final decisions.", |
| "Review rating distributions are biased toward accepted papers for NeurIPS 2025 (95.4% of the dataset). Rating statistics for this venue reflect reviewer behavior on accepted papers only and are not representative of reviews on rejected submissions.", |
| "ICLR 2024 reviews have empty Rating fields (they use text-based Soundness/Presentation/Contribution fields instead), which limits quantitative rating analysis for this venue.", |
| "ICML 2025 reviews use a different format (Overall Recommendation) without the same structured fields as ICLR/NeurIPS reviews, making cross-venue review analysis difficult.", |
| "OpenReview-sourced data (ICLR) may have self-selection bias, not all authors opt to make their reviews public.", |
| "The conferences represented are all top-tier ML venues, so the dataset reflects cutting-edge ML research trends and may not generalize to other CS subfields or lower-tier venues.", |
| "English-language papers are overwhelmingly represented, potentially underrepresenting non-English ML research communities." |
| ], |
| "rai:personalSensitiveInformation": "The dataset contains publicly available academic papers and their peer reviews. Author names and institutional affiliations are included as part of standard academic metadata. Review text is sourced from OpenReview (ICLR) where reviews are public. No private personal information such as emails, addresses, or demographic data about authors or reviewers is included beyond what is publicly available on the source platforms.", |
| "rai:dataUseCases": [ |
| "Acceptance/Rejection Prediction - Predict paper acceptance from text and review scores. Valid only for ICLR venues (2024-2026) which have balanced accept/reject coverage.", |
| "Peer Review Analysis - Analyze reviewer behavior, detect bias, and study review quality patterns. Valid only for ICLR venues.", |
| "Paper Quality Prediction - Predict paper impact or quality from text features. Valid for all venues.", |
| "Citation Network Analysis - Build citation graphs from bibliography data. Valid for all venues.", |
| "NLP for Scientific Text - Train and evaluate models on scientific document understanding, summarization, and information extraction. Valid for all venues.", |
| "Argument Mining - Extract strengths, weaknesses, and arguments from reviews. Valid only for ICLR venues.", |
| "Meta-Science / Research Trends - Study trends in ML research, topic modeling, method evolution. Valid for all venues.", |
| "Use cases NOT validated and NOT recommended: (1) cross-venue rating comparison (different scales and formats across venues); (2) studying NeurIPS 2025 or ICML 2025 rejection patterns (selection bias — mostly accepted papers); (3) as a representative sample of all CS research (top-tier ML only); (4) multilingual NLP research (English-language papers are overwhelmingly represented); (5) training models for clinical or medical decision support; (6) causal claims about peer review quality or fairness (observational data only, no experimental control)." |
| ], |
| "rai:dataSocialImpact": "This dataset can advance meta-science research by enabling large-scale analysis of the peer review process, potentially leading to improvements in review quality and fairness. Positive impacts include: enabling transparency research in academic publishing, supporting development of tools to assist reviewers, and facilitating study of research trends. Negative risks include: potential for gaming the review system if models predict reviewer preferences, privacy concerns if reviewer identities can be inferred from review patterns, and reinforcement of existing biases in top-tier venue representation if used uncritically for broad claims about ML research.", |
| "rai:hasSyntheticData": false, |
| "prov:wasDerivedFrom": [ |
| { |
| "@id": "https://openreview.net", |
| "prov:label": "OpenReview", |
| "description": "OpenReview platform hosting peer reviews, decisions, and paper metadata for ICLR 2024, ICLR 2025, and ICLR 2026. All submissions (accepted, rejected, and pending) are included for ICLR venues.", |
| "sc:license": "https://creativecommons.org/licenses/by/4.0/", |
| "prov:wasAttributedTo": { |
| "@id": "openreview_foundation", |
| "prov:label": "OpenReview.net Foundation" |
| } |
| }, |
| { |
| "@id": "https://icml.cc/Conferences/2025", |
| "prov:label": "ICML 2025 Conference Proceedings", |
| "description": "Accepted paper list and metadata from the International Conference on Machine Learning 2025. Only accepted papers (~26.9% of total submissions) are included; rejected papers are not available from this source.", |
| "sc:license": "https://creativecommons.org/licenses/by/4.0/", |
| "prov:wasAttributedTo": { |
| "@id": "imls", |
| "prov:label": "International Machine Learning Society (IMLS)" |
| } |
| }, |
| { |
| "@id": "https://neurips.cc/Conferences/2025", |
| "prov:label": "NeurIPS 2025 Conference Proceedings", |
| "description": "Accepted paper list and metadata from the Conference on Neural Information Processing Systems 2025. Only accepted papers (~24.5% of total submissions) are included; rejected papers are not available from this source. NeurIPS states that authors retain copyright and grant NeurIPS a non-exclusive publication/distribution license.", |
| "sc:license": "https://nips.cc/FAQ/Copyright", |
| "prov:wasAttributedTo": { |
| "@id": "neurips_foundation", |
| "prov:label": "NeurIPS Foundation" |
| } |
| } |
| ], |
| "prov:wasGeneratedBy": [ |
| { |
| "@type": "prov:Activity", |
| "prov:type": { |
| "@id": "https://www.wikidata.org/wiki/Q4929239" |
| }, |
| "prov:label": "Web Scraping: OpenReview and Conference Proceedings", |
| "prov:atTime": "2026-03-01T00:00:00Z", |
| "description": "Paper metadata, reviews, decisions, and PDF links were scraped from the web. For ICLR 2024, 2025, and 2026, data was collected from OpenReview (https://openreview.net) using web scraping and the OpenReview Python API. The ICLR venues include all submissions — accepted, rejected, and pending papers. For ICLR venues, scraping occurred during the review period — after initial peer reviews were posted but before meta-reviews and final decisions were announced — resulting in ~20-27% of papers having 'Pending' as their decision. For ICML 2025 and NeurIPS 2025, data was collected from the official conference proceedings pages, which primarily include accepted papers only (~26.9% and ~24.5% of total submissions respectively). Collection period: September 2025 to March 2026.", |
| "prov:wasAttributedTo": [ |
| { |
| "@type": "prov:Agent", |
| "@id": "research_team_scraping", |
| "prov:label": "Anonymous Research Team", |
| "prov:description": "Designed and executed the web scraping pipeline, wrote custom crawlers for OpenReview and conference proceedings pages." |
| }, |
| { |
| "@type": "prov:SoftwareAgent", |
| "@id": "openreview_api", |
| "prov:label": "OpenReview Python API", |
| "description": "Official OpenReview Python client (openreview-py) and REST API for programmatic access to paper metadata, reviews, and decisions from ICLR 2024, 2025, and 2026." |
| } |
| ] |
| }, |
| { |
| "@type": "prov:Activity", |
| "prov:type": { |
| "@id": "https://www.wikidata.org/wiki/Q1166228" |
| }, |
| "prov:label": "PDF Processing with GROBID", |
| "prov:atTime": "2026-03-15T00:00:00Z", |
| "description": "Downloaded PDFs were processed through GROBID (https://github.com/kermitt2/grobid), a machine-learning-based document parser for academic papers. GROBID extracts structured metadata (title, authors, abstract, keywords, date), bibliography in both JSON and BibTeX formats, full TEI XML document structure (sections, figures, tables, equations), and clean full-text extraction with section boundaries. Review ratings from the scraped JSON data were parsed and normalized where possible (e.g., extracting numeric values from strings like '3: reject, not good enough').", |
| "prov:wasAttributedTo": [ |
| { |
| "@type": "prov:Agent", |
| "@id": "research_team_grobid", |
| "prov:label": "Anonymous Research Team", |
| "prov:description": "Designed and executed the web scraping pipeline, wrote custom crawlers for OpenReview and conference proceedings pages." |
| }, |
| { |
| "@type": "prov:SoftwareAgent", |
| "@id": "grobid", |
| "prov:label": "GROBID v0.8.1", |
| "description": "Machine-learning-based PDF document parser for extracting structured metadata, bibliography, and full text from academic papers (https://github.com/kermitt2/grobid)." |
| } |
| ] |
| }, |
| { |
| "@type": "prov:Activity", |
| "prov:type": { |
| "@id": "https://www.wikidata.org/wiki/Q21310498" |
| }, |
| "prov:label": "Data Consolidation to Apache Parquet", |
| "prov:atTime": "2026-04-01T00:00:00Z", |
| "description": "All structured data from the scraping and GROBID processing steps was consolidated into a single Apache Parquet file (papers.parquet, 1.8 GB) with zstd compression using PyArrow. Long text fields (full_text, grobid_fulltext, meta_review, bibliography) are stored as strings; reviews are stored as JSON strings due to their variable-length nested structure. The result is a single queryable table with 47,214 rows and 31 columns.", |
| "prov:wasAttributedTo": [ |
| { |
| "@type": "prov:Agent", |
| "@id": "research_team_consolidation", |
| "prov:label": "Anonymous Research Team", |
| "prov:description": "Wrote the consolidation script, defined the schema, and validated the output." |
| }, |
| { |
| "@type": "prov:SoftwareAgent", |
| "@id": "pyarrow", |
| "prov:label": "PyArrow (Apache Arrow)", |
| "description": "Apache Arrow Python library (PyArrow) used to write the consolidated Parquet file with zstd compression and schema definition." |
| } |
| ] |
| } |
| ], |
| "distribution": [ |
| { |
| "@type": "cr:FileObject", |
| "@id": "papers_parquet", |
| "name": "papers.parquet", |
| "description": "Combined tabular dataset containing all 47,214 papers across 5 venues with 31 columns. Apache Parquet format with zstd compression (1.8 GB).", |
| "contentUrl": "https://huggingface.co/datasets/anoyresearcher/prism_paper_data/resolve/main/papers.parquet", |
| "encodingFormat": "application/x-parquet", |
| "sha256": "96838db82e79ff1baecf7b5a57107847a2eba702f4d33128dfff29277dc14766", |
| "contentSize": "1926458879" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "iclr_2024_zip", |
| "name": "ICLR_2024.zip", |
| "description": "ICLR 2024 papers in file-based format (json, txt, grobid outputs). 7,262 papers. Does not include PDFs.", |
| "contentUrl": "https://huggingface.co/datasets/anoyresearcher/prism_paper_data/resolve/main/ICLR_2024.zip", |
| "encodingFormat": "application/zip", |
| "sha256": "d407f8b93b1dfab763763feff09a477d44ada0711eee84d88af5e70db40ae14e", |
| "contentSize": "722410270" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "iclr_2025_zip", |
| "name": "ICLR_2025.zip", |
| "description": "ICLR 2025 papers in file-based format (json, txt, grobid outputs). 11,519 papers. Does not include PDFs.", |
| "contentUrl": "https://huggingface.co/datasets/anoyresearcher/prism_paper_data/resolve/main/ICLR_2025.zip", |
| "encodingFormat": "application/zip", |
| "sha256": "63e5e0812fddc1448a5ed46e22ae3525ff8a65262effd4bc6571b81305dea48e", |
| "contentSize": "805125145" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "iclr_2026_zip", |
| "name": "ICLR_2026.zip", |
| "description": "ICLR 2026 papers in file-based format (json, txt, grobid outputs). 19,471 papers. Does not include PDFs.", |
| "contentUrl": "https://huggingface.co/datasets/anoyresearcher/prism_paper_data/resolve/main/ICLR_2026.zip", |
| "encodingFormat": "application/zip", |
| "sha256": "4858c329b8fedf8072a1f644ffbd42f3e99383eacf825a32b9920b3e4a291a25", |
| "contentSize": "1443696594" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "icml_2025_zip", |
| "name": "ICML_2025.zip", |
| "description": "ICML 2025 papers in file-based format (json, txt, grobid outputs). 3,422 papers. Does not include PDFs.", |
| "contentUrl": "https://huggingface.co/datasets/anoyresearcher/prism_paper_data/resolve/main/ICML_2025.zip", |
| "encodingFormat": "application/zip", |
| "sha256": "f4914c26b2a4c46f1e9d540388c53321a505712eab870d2b2248adcf40d535c5", |
| "contentSize": "251652800" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "neurips_2025_zip", |
| "name": "NeurIPS_2025.zip", |
| "description": "NeurIPS 2025 papers in file-based format (json, txt, grobid outputs). 5,540 papers. Does not include PDFs.", |
| "contentUrl": "https://huggingface.co/datasets/anoyresearcher/prism_paper_data/resolve/main/NeurIPS_2025.zip", |
| "encodingFormat": "application/zip", |
| "sha256": "dc0b083fa96a0bcd68284ddbffc3b888b3cdde76bdfb5be4edd72f0939b9ddbc", |
| "contentSize": "485442851" |
| } |
| ], |
| "recordSet": [ |
| { |
| "@type": "cr:RecordSet", |
| "@id": "papers", |
| "name": "papers", |
| "description": "Main record set containing all 47,214 ML conference papers with review data, GROBID metadata, and full text.", |
| "key": ["venue", "paper_id"], |
| "field": [ |
| { |
| "@type": "cr:Field", |
| "@id": "papers/paper_id", |
| "name": "paper_id", |
| "description": "Source-specific paper ID. For OpenReview-sourced venues this is the OpenReview paper ID (e.g., 00ezkB2iZf); for proceedings-sourced venues this is the dataset's source paper identifier. Use together with venue as the dataset key.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "paper_id" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/venue", |
| "name": "venue", |
| "description": "Conference venue. One of: ICLR_2024, ICLR_2025, ICLR_2026, ICML_2025, NeurIPS_2025.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "venue" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/decision", |
| "name": "decision", |
| "description": "Editorial decision. Values: Accept (poster), Accept (spotlight), Accept (oral), Conditional Accept, Reject, Pending. Note: ICML/NeurIPS data is biased toward accepted papers.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "decision" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/meta_review", |
| "name": "meta_review", |
| "description": "Meta-review text from the area chair.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "meta_review" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/num_reviews", |
| "name": "num_reviews", |
| "description": "Number of peer reviews for this paper.", |
| "dataType": "sc:Integer", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "num_reviews" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/rating_avg", |
| "name": "rating_avg", |
| "description": "Average reviewer rating (where available). Rating scales differ by venue.", |
| "dataType": "sc:Float", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "rating_avg" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/rating_min", |
| "name": "rating_min", |
| "description": "Minimum reviewer rating.", |
| "dataType": "sc:Integer", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "rating_min" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/rating_max", |
| "name": "rating_max", |
| "description": "Maximum reviewer rating.", |
| "dataType": "sc:Integer", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "rating_max" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/confidence_avg", |
| "name": "confidence_avg", |
| "description": "Average reviewer confidence score.", |
| "dataType": "sc:Float", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "confidence_avg" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/soundness_avg", |
| "name": "soundness_avg", |
| "description": "Average soundness score (ICLR 2024 uses poor/fair/good/excellent scale).", |
| "dataType": "sc:Float", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "soundness_avg" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/presentation_avg", |
| "name": "presentation_avg", |
| "description": "Average presentation score.", |
| "dataType": "sc:Float", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "presentation_avg" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/contribution_avg", |
| "name": "contribution_avg", |
| "description": "Average contribution score.", |
| "dataType": "sc:Float", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "contribution_avg" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/reviews_json", |
| "name": "reviews_json", |
| "description": "Full peer reviews as a JSON string array. Each review contains fields like Rating, Confidence, Summary, Strengths, Weaknesses, Questions, Limitations. Field names and scales differ between venues.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "reviews_json" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/keywords", |
| "name": "keywords", |
| "description": "Paper keywords as a JSON array string.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "keywords" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/primary_area", |
| "name": "primary_area", |
| "description": "Primary subject area of the paper (e.g., 'Safety in Machine Learning').", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "primary_area" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/subject_areas", |
| "name": "subject_areas", |
| "description": "Subject areas as a JSON array string.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "subject_areas" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/review_title", |
| "name": "review_title", |
| "description": "Paper title from the submission metadata (available for all ICLR; may be empty for ICML/NeurIPS).", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "review_title" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/review_abstract", |
| "name": "review_abstract", |
| "description": "Abstract from the submission metadata (available for all ICLR; may be empty for ICML/NeurIPS).", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "review_abstract" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/grobid_title", |
| "name": "grobid_title", |
| "description": "Title extracted by GROBID from the PDF.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "grobid_title" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/grobid_abstract", |
| "name": "grobid_abstract", |
| "description": "Abstract extracted by GROBID from the PDF.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "grobid_abstract" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/grobid_authors", |
| "name": "grobid_authors", |
| "description": "Authors extracted by GROBID as a JSON array string.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "grobid_authors" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/grobid_keywords", |
| "name": "grobid_keywords", |
| "description": "Keywords extracted by GROBID as a JSON array string.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "grobid_keywords" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/grobid_date", |
| "name": "grobid_date", |
| "description": "Publication/acceptance date extracted by GROBID.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "grobid_date" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/full_text", |
| "name": "full_text", |
| "description": "Full paper text extracted from PDF (plain text, may contain OCR artifacts).", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "full_text" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/grobid_fulltext", |
| "name": "grobid_fulltext", |
| "description": "GROBID-extracted full text (cleaner than full_text, preserves section structure).", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "grobid_fulltext" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/bibliography_json", |
| "name": "bibliography_json", |
| "description": "Bibliography references as a JSON array string. Each reference has title, authors, year, venue fields.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "bibliography_json" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/bibliography_bib", |
| "name": "bibliography_bib", |
| "description": "Bibliography in BibTeX format.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "bibliography_bib" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/pdf_path", |
| "name": "pdf_path", |
| "description": "Relative path to the original PDF file (PDFs are not included in the dataset distribution).", |
| "dataType": "sc:Text", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "pdf_path" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/stat_num_reviews", |
| "name": "stat_num_reviews", |
| "description": "Number of reviews from the statistics field in the review JSON.", |
| "dataType": "sc:Integer", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "stat_num_reviews" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/stat_has_meta_review", |
| "name": "stat_has_meta_review", |
| "description": "Whether a meta-review exists for this paper.", |
| "dataType": "sc:Boolean", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "stat_has_meta_review" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "papers/stat_has_decision", |
| "name": "stat_has_decision", |
| "description": "Whether a decision exists for this paper.", |
| "dataType": "sc:Boolean", |
| "source": { |
| "fileObject": { |
| "@id": "papers_parquet" |
| }, |
| "extract": { |
| "column": "stat_has_decision" |
| } |
| } |
| } |
| ] |
| } |
| ] |
| } |
|
|
