srm-ist/SRM-Publications-Dataset_2019-2024

📚 SRM Publications Dataset (2019-2024)

Instruction-Tuning Dataset for Academic Research & Publication Retrieval

---

🎯 Dataset Description

This dataset contains publication metadata from SRM Institute of Science and Technology spanning 2019-2024, formatted for instruction-following and question-answering tasks.

🔢 22,809 unique publications → 933,010 instruction-response pairs across 45 diverse instruction types

Perfect for training AI models to understand and retrieve academic publication information!

---

🌐 Source

📖 Original Data: SRM Institute of Science and Technology - Research Publications

🔍 Coverage: Scopus-indexed publications from SRM Institute researchers (2019-2024)

---

📊 Dataset Structure

🔑 Data Fields

Each entry contains four fields following the Alpaca instruction format:

Field	Description
💬 instruction	The task or question to be performed (45 unique types)
📥 input	Context or parameters (paper title, DOI, authors, etc.)
📤 output	The expected response or answer
📝 text	Formatted Alpaca-style string (instruction + input + output)

💡 Example Entry

{
  "instruction": "Who are the authors of this research paper?",
  "input": "Title: Controlling the electronic and magnetic properties of Mn2CoAl by d metal doping: A DFT study\nDOI: 10.1016/j.jpcs.2024.111914",
  "output": "Karunakaran M.; Ravi D.; Bandyopadhyay P.; Banerjee R.",
  "text": "### Instruction:\nWho are the authors of this research paper?\n\n### Input:\nTitle: Controlling the electronic and magnetic properties of Mn2CoAl by d metal doping: A DFT study\nDOI: 10.1016/j.jpcs.2024.111914\n\n### Response:\nKarunakaran M.; Ravi D.; Bandyopadhyay P.; Banerjee R."
}

📈 Data Splits

Split	📊 Entries	📐 Percentage	💾 Size
🏋️ Train	746,408	80%	389 MB
✅ Validation	93,301	10%	49 MB
🧪 Test	93,301	10%	49 MB
🎯 Total	933,010	100%	487 MB

---

🎨 Instruction Types (45 Categories)

The dataset includes diverse instruction types across 10 major categories:

🔗 1. DOI Queries (4 types)

• Finding DOI from publication metadata
• Retrieving publication details from DOI
• Looking up paper titles from DOI
• Cross-referencing DOI with other metadata

👥 2. Author Queries (6 types)

• Identifying authors of papers
• Finding publications by specific authors
• Listing research team members
• Author contribution analysis
• Research output tracking
• Team collaboration discovery

📊 3. Citation Queries (5 types)

• Citation counts and metrics
• Research impact assessment
• Citation-based comparisons
• Influence evaluation
• Impact factor analysis

📰 4. Source/Journal Queries (6 types)

• Publication venue identification
• Journal/conference searches
• Venue-based paper discovery
• Source title lookups
• Publication outlet analysis
• Venue-specific research trends

📅 5. Temporal Queries (5 types)

• Publication year identification
• Year-based paper searches
• Temporal research trends
• Historical publication tracking
• Time-series analysis

📖 6. Volume/Issue Queries (3 types)

• Volume and issue identification
• Issue-specific paper searches
• Journal navigation

📄 7. Document Type Queries (4 types)

• Publication type classification
• Type-based filtering
• Article vs. conference paper distinction
• Document category analysis

🔓 8. Open Access Queries (3 types)

• Access status verification
• Open access paper discovery
• Accessibility checking

📚 9. Comprehensive Metadata (5 types)

• Complete bibliographic information
• Citation entry generation
• Full publication details
• Metadata extraction
• Summary generation

🔍 10. Search/Discovery Queries (4 types)

• Topic-based searches
• Multi-criteria discovery
• Related research finding
• Parameter-based location

---

📋 Metadata Coverage

Publications include the following metadata fields (when available):

🏷️ Field	📝 Description
👤 Authors	Author names and IDs
📌 Title	Publication title
📆 Year	Publication year
🔗 DOI	Digital Object Identifier
📰 Source	Journal/conference name
📖 Volume/Issue	Volume and issue numbers
📊 Citations	Citation counts
📄 Document Type	Article, conference paper, etc.
🔓 Open Access	Access status
📃 Page Info	Article numbers and page ranges

---

🚀 Use Cases

This dataset is perfect for:

🤖 Model Training ✨ Instruction-following model fine-tuning 🎯 Academic search & retrieval systems 💬 Question-answering models 🧠 LLM specialization for research

🔬 Research Applications 📚 Bibliographic information extraction 🔍 Research discovery tools 📊 Citation analysis systems 🎓 Academic assistant development

---

✅ Data Quality

Metric	Status	Details
🎯 Empty Outputs	✅ 0%	All entries contain valid responses
⚖️ Distribution	✅ Balanced	Each instruction: 0.6-2.5% of dataset
🔄 Duplicates	✅ None	Proper train/val/test separation
📐 Formatting	✅ Clean	Consistent structure across all entries

---

⚠️ Limitations

Important considerations when using this dataset:

• 🏛️ Data is specific to SRM Institute of Science and Technology publications
• 📅 Limited to Scopus-indexed publications (2019-2024)
• 📝 Some metadata fields may be incomplete (abstracts/keywords not in all records)
• 📊 Citation counts are from data collection time and may be outdated
• 🌍 Geographic and institutional scope is limited to one institution

---

📖 Citation

If you use this dataset, please cite:

@dataset{srm_publications_2024,
  title={SRM Publications Dataset (2019-2024)},
  author={Anuraag Rath},
  year={2026},
  url={https://www.srmist.edu.in/research/publications/},
  note={Instruction-tuning dataset with 933K examples from 22.8K publications}
}

---

🛠️ Dataset Creation

Click to see creation methodology

This dataset was created through the following pipeline:

1. 📥 Data Collection: Extracted publication metadata from SRM Institute's Scopus-indexed publications (2019-2024)
2. 🔄 Transformation: Generated diverse instruction-response pairs from each publication record
3. 🎨 Diversification: Created 45 unique instruction types across 10 categories
4. ✂️ Splitting: Divided data into train/validation/test sets (80/10/10)
5. 📝 Formatting: Structured in Alpaca instruction format for LLM training

Tools Used: Python, xlrd, pandas

---

📌 Version Info

🏷️ Version	📅 Release Date	📊 Coverage	📚 Publications	🎯 Examples
1.0	April 2026	2019-2024	22,809	933,010

---

🌟 Made with ❤️ for the Research Community

Happy Training! 🚀

👨‍💻 Created by

Do follow me for more 😉

---

🙏 Please upvote ⬆️ this dataset if you find this to be useful

---