InfoBayAI/English-Thesis-Dataset

Thesis Dataset

Dataset Description

The Thesis Dataset is a large-scale academic text corpus designed for building next-generation Natural Language Processing (NLP) systems, Large Language Models (LLMs), Retrieval-Augmented Generation (RAG) systems, research assistants, document understanding models, and knowledge extraction pipelines.

The complete multilingual collection contains over 627,000 thesis documents comprising 7.38+ billion words. This English release contains 512,129 English thesis documents, representing approximately 6.02 billion words of scholarly content.

Each thesis contains structured academic writing including:

• Abstracts
• Literature Reviews
• Research Methodologies
• Experimental Results
• Discussions
• Conclusions
• Technical Explanations
• Academic References
• Tables
• Figures and Diagrams

The dataset is suitable for large-scale language modeling, academic reasoning, long-context understanding, document summarization, question answering, citation-aware AI, and educational research.

---

Dataset Statistics

Metric	English Dataset
Original Thesis Documents	512,129
After Initial Quality Filtering (20%)	409,703
High-Quality Usable Thesis (42%)	172,075
Average Pages per Thesis	140
Average Words per Page	250
Total Pages	24,090,548
Total Word Count	6.02 Billion Words

---

Overall Multilingual Collection

Language	Thesis	Word Count
English	512,129	6.02B
Hindi	50,013	0.59B
Other Languages	65,170	0.77B
Total	627,312	7.38B+ Words

---

Dataset Specification

Language: English

Document Type:
Academic Thesis

Original Documents:
512,129

Filtered High-Quality Thesis:
172,075

Average Pages:
140

Average Words per Page:
250

Total Pages:
24,090,548

Total Words:
6.02 Billion

Overall Multilingual Collection:
627,312 Thesis
7.38+ Billion Words

---

Dataset Features

• Large-scale academic corpus
• Long-form structured documents
• Rich technical language
• Multiple academic disciplines
• Chapter-level organization
• Figures and diagrams
• Real-world scholarly writing
• Research-grade textual quality
• Suitable for long-context LLM training

---

Key Use Cases

Large Language Model (LLM) Pretraining

Instruction Tuning

Supervised Fine-Tuning (SFT)

Self-Supervised Learning (SSL)

Reinforcement Learning from Human Feedback (RLHF)

Retrieval-Augmented Generation (RAG)

Academic Question Answering

Document Understanding

Research Assistants

Knowledge Extraction

Citation Analysis

Scientific Summarization

Educational AI

Information Retrieval

Long Document Modeling

Domain-specific Language Modeling

Benchmark Creation

---

Value of This Dataset

This dataset enables researchers and organizations to build production-ready AI systems capable of understanding complex scholarly documents.

Applications include:

• Academic Copilots
• Research Search Engines
• Literature Review Automation
• Knowledge Graph Construction
• Scientific Question Answering
• Long-Document Retrieval
• AI Tutors
• Educational Search Systems
• Enterprise Research Intelligence
• Document Summarization Systems

The combination of billions of words and long-form academic structure makes this dataset particularly valuable for training foundation models requiring extensive contextual understanding.

---

Basic JSON Schema

{
  "pdf_name": "string",
  "page_no": "int32",
  "relevant_chapter": "string",
  "theory": "string",
  "diagram_image": "Image"
}

---

Data Creation

The dataset has been procured through formal agreements and generated in the ordinary course of business.

Quality processing includes:

• Initial document quality filtering
• Duplicate reduction
• Removal of low-quality content
• Exclusion of partially copied material
• Academic structure preservation

Statistics are estimated using an average of 140 pages per thesis and 250 words per page.

---

Considerations

This repository contains sample data intended for research and educational purposes.

The complete multilingual collection consists of:

• 627,312 Thesis Documents
• 7.38+ Billion Words
• English
• Hindi
• Additional Languages

Enterprise licensing and access to the complete dataset are available through InfoBay.AI.

Website: https://infobay.ai/

Email: datareq@infobay.ai

Phone: (+91) 8303174762

---

Citation

If you use this dataset in your research, publications, or commercial applications, please cite the corresponding dataset repository.

---

License

CC BY 4.0