Name: R&D and AI product dataset
Creator: Changsha Langhui Information Technology Co., Ltd.
Keywords: R&D, AI products, paper data, patent data, literature search

Dataset Overview

Dataset Name	Technology R&D and AI Product Dataset
Total Data Volume	65,000 (paper abstracts + patents + technical solutions + product documents)
Coverage	Artificial Intelligence / Big Data / Cloud Computing / Networking / Blockchain / Quantum Computing / New Materials
Time Span	2018 - 2026
Subset Composition	Academic paper abstracts 25K / Patent text 18K / Technical programs 12K / Product documents 10K
Data Source	ArXiv / IEEE / CNKI / National Intellectual Property Administration / Open Business Technical Documentation
Languages	Bilingual Chinese-English (Chinese 58% / English 42%)
Annotation Method	+ NLP Auto-Classification Supplement + Intellectual Property Analysis

Detailed Description of Data Dimensions

Field Name	Data Type	Value Range / Description
doc_id	String(32)	Document_UNIQUE_TECH__Evil _UNIQUE_TECH___
doc_type	Enum	PAPER \| PATENT \| TECH_SOLUTION \| PRODUCT_DOC
domain	Enum	AI \| BIG_DATA \| CLOUD \| IOT \| BLOCKCHAIN \| QUANTUM \| NEW_MATERIAL
subdomain	String	Subfield labels (e.g., NLP / CV / Reinforcement Learning / Edge Computing)
title	Text	Title of the paper/patent/document
abstract	Text	Abstract (originally in Chinese)
language	Enum	ZH \| EN
publish_year	Integer	Year published
keywords	Array	Keyword list (Chinese-English standardization)
tech_methods	Array	Technical methods involved (e.g., Transformer / GAN / Diffusion)
benchmark_datasets	Array	Reference benchmark datasets (e.g., ImageNet / COCO / SQuAD)
citation_count	Integer	Number of citations (papers only)
patent_status	Enum	APPLIED · GRANTED · EXPIRED (patents only)
innovation_level	Enum	INCREMENTAL \| BREAKTHROUGH \| DISRUPTIVE
trl_level	Integer	TrL 1-9

De-identified JSON Data Sample

[ { "doc_id": "RND-PAPER-2024-AI-05231", "doc_type": "PAPER", "domain": "AI", "subdomain": "NLP", "title": "Large Language Models Are Zero-Shot Reasoners", "abstract": "We demonstrate that LLMs are decent zero-shot reasoners by simply adding 'Let''s think step by step' before each answer...", "language": "EN", "publish_year": 2024, "keywords": ["[]LanguageModels", "[]Sample[]", "[chain-of-thought]", "[]"], "tech_methods": ["Chain-of-Thought", "Zero-shot Prompting", "LLM"], "benchmark_datasets": ["GSM8K", "SVAMP", "AQuA", "StrategyQA"], "citation_count": 1250, "patent_status": null, "innovation_level": "BREAKTHROUGH", "trl_level": null }, { "doc_id": "RND-PATENT-2023-CLOUD-01892", "doc_type": "PATENT", "domain": "CLOUD", "subdomain": "[]", "title": "dataReal-timeProcessingLaw", "abstract": "dataReal-timeProcessingLaw，AIModels...", "language": "ZH", "publish_year": 2023, "keywords": ["[]", "[]", "[]", "Real-timeProcessing"], "tech_methods": ["[]", "[]", "Models[]"], "benchmark_datasets": [], "citation_count": null, "patent_status": "GRANTED", "innovation_level": "INCREMENTAL", "trl_level": 7 }, { "doc_id": "RND-SOLUTION-2025-IOT-00765", "doc_type": "TECH_SOLUTION", "domain": "IOT", "subdomain": "[]", "title": "[[]Enterprise][]Solutions[]", "abstract": "Matterplatform，AIScenes...", "language": "ZH", "publish_year": 2025, "keywords": ["[]", "Matter[]", "AIScenes", "[]"], "tech_methods": ["Matter", "[]AI", "Scenes[]"], "benchmark_datasets": [], "citation_count": null, "patent_status": null, "innovation_level": "INCREMENTAL", "trl_level": 8 }, { "doc_id": "RND-PAPER-2024-QUANTUM-00342", "doc_type": "PAPER", "domain": "QUANTUM", "subdomain": "[]", "title": "Quantum Neural Networks for Financial Time Series Prediction", "abstract": "We propose a hybrid quantum-classical neural network architecture that leverages quantum feature maps for financial time series forecasting...", "language": "EN", "publish_year": 2024, "keywords": ["[]", "[]", "[]", "[finance]"], "tech_methods": ["Quantum Neural Network", "Variational Quantum Circuit", "Hybrid Classical-Quantum"], "benchmark_datasets": ["S&P500", "NASDAQ"], "citation_count": 85, "patent_status": null, "innovation_level": "BREAKTHROUGH", "trl_level": 4 }, { "doc_id": "RND-PRODUCT-2025-AI-00987", "doc_type": "PRODUCT_DOC", "domain": "AI", "subdomain": "MultimodularAI", "title": "Gemini 2.0 MultimodalLarge Models[]", "abstract": "Gemini 2.0Multimodular，Multimodular、 and Processing...", "language": "ZH", "publish_year": 2025, "keywords": ["Multimodular", "[]LanguageModels", "Gemini", "Transformer"], "tech_methods": ["MoE", "Multi-modal Encoder", "Cross-attention", "Long Context"], "benchmark_datasets": ["MMLU", "MMBench", "Video-MME"], "citation_count": null, "patent_status": null, "innovation_level": "DISRUPTIVE", "trl_level": 9 } ]

Annotation Specifications and Quality Control

Annotation Standards

Field classification follows the ACM Computing Classification System (CS 2012) taxonomy
Technical methods are tagged with standardized technical terminology covering 12 major categories and 200+ methods, including AI / systems / networking / security.
TRL (Technology Readiness Level) assessed by domain experts according to the NASA/DoD TRL 9 scale.
Innovation level is annotated by two annotators, with a Kappa consistency coefficient > 0.85.
Chinese-English keyword alignment uses a custom glossary to ensure consistency in cross-language retrieval.

Quality Metrics

97.1%

Domain classification accuracy

94.3%

Technical method label F1

0.87

Innovation level Kappa

99.2%

Bilingual keyword alignment rate

AI Training Use Cases

Intelligent Search and Recommendation of Scientific and Technical Literature

Applied models: Sentence-BERT + dual-tower recall / ColBERT. Trains a semantic search engine for scientific literature based on document abstracts and keywords.

Novelty and infringement analysis for patent examination

Applied models: Patent-BERT + contrastive learning. Trains a patent similarity matching model to automatically identify potential infringement risks and technology gaps.

Technology trend forecasting

Applied models: LDA / BERTopic + time series analysis. Identifies emerging technology topics from large volumes of scientific and technical literature and forecasts trends.

Automated technical proposal writing

Applied models: LLM (GPT-4o / DeepSeek-V3) + RAG. Automatically generates technical proposals and feasibility reports based on the domain knowledge base and patent repository.

Get This Dataset

R&D datasets are flexibly authorized by domain and document type.

137-5502-0164 Online Consultation

Related Datasets

Commercial Application Value

Enterprise R&D Intelligence

Provides technology monitoring and competitive analysis capabilities to R&D teams in science and technology enterprises, reduces the risk of duplicated R&D, and shortens the technology research cycle by 50%.

Intellectual Property Services

AI-assisted patent retrieval and infringement analysis tools provided to patent agencies and law firms have tripled the efficiency of patent drafting and review.

Academic Research Support

Automated literature synthesis, research hotspot tracking, and cross-disciplinary analysis capabilities for universities and research institutions.

Request technology research and AI product data

Changsha Langhui Information Technology Co., Ltd. The DataSetsAPI platform is dedicated to providing AI enterprises and research institutions with high-quality, compliant data assets.

Data Sample Preview · Customized Data Solutions · Professional Technical Support

Data Sample Preview Customized Discipline-Specific Solutions Professional Technical Support

Technology R&D and AI Product Dataset