RohitManglik commited on
Commit
349bb76
·
verified ·
1 Parent(s): 0a0b38a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +313 -0
README.md CHANGED
@@ -1,3 +1,316 @@
1
  ---
2
  license: cc-by-4.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: cc-by-4.0
3
+ language:
4
+ - en
5
+ - hi
6
+ - te
7
+ - mr
8
+ - bn
9
+ - ar
10
+ - ta
11
+ - gu
12
+ - kn
13
+ - pa
14
+ - or
15
+ - ml
16
+
17
+ pretty_name: Multilingual STEM & Non-STEM Question Answering Dataset
18
+ size_categories:
19
+ - 1M<n<10M
20
+
21
+ tags:
22
+ - question-answering
23
+ - educational
24
+ - stem
25
+ - non-stem
26
+ - multilingual
27
+ - exam-preparation
28
+ - reasoning
29
+ - mathematics
30
+ - science
31
+ - engineering
32
+ - medical
33
+ - social-science
34
+ - competitive-exams
35
+ - llm
36
+ - instruction-tuning
37
+ - synthetic-data
38
+ - rag
39
+ - pdf
40
+ - json
41
+ - education
42
  ---
43
+
44
+ # Multilingual STEM & Non-STEM Question Answering Dataset (6.7M+ Q&A)
45
+
46
+ ## Overview
47
+
48
+ The **Multilingual STEM & Non-STEM Question Answering Dataset** is one of the largest educational Q&A resources designed for AI, Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), educational assistants, intelligent tutoring systems, and academic research.
49
+
50
+ The dataset contains **over 6.7 million educational question-answer pairs** spanning STEM, Non-STEM, competitive examinations, language learning, professional education, and general knowledge domains.
51
+
52
+ Every sample includes a **question, multiple-choice options, correct answer, and detailed explanation**, making the dataset suitable for both supervised fine-tuning and reasoning-focused AI systems.
53
+
54
+ The complete collection is available in both **PDF** and **structured JSON** formats.
55
+
56
+ ---
57
+
58
+ # Dataset Highlights
59
+
60
+ | Property | Value |
61
+ |-----------|--------|
62
+ | Total Questions | **6.7M+** |
63
+ | English Questions | **2.5M+** |
64
+ | Total Tokens | **1.8B+** |
65
+ | Average Words per Question | **~210** |
66
+ | Formats | PDF, JSON |
67
+ | Languages | 12 |
68
+ | Domains | STEM + Non-STEM |
69
+ | Question Type | Multiple Choice |
70
+ | Includes Explanations | ✅ |
71
+ | Mathematical Equations | LaTeX & MathML |
72
+ | Embedded Educational Images | ✅ |
73
+
74
+ ---
75
+
76
+ # Supported Languages
77
+
78
+ | Language | Questions |
79
+ |-----------|-----------:|
80
+ | English | 4,073,358 |
81
+ | Hindi | 1,645,977 |
82
+ | Telugu | 158,044 |
83
+ | Bengali | 145,979 |
84
+ | Arabic | 139,496 |
85
+ | Marathi | 135,971 |
86
+ | Tamil | 102,569 |
87
+ | Gujarati | 69,580 |
88
+ | Kannada | 56,976 |
89
+ | Punjabi | 20,114 |
90
+ | Odia | 18,465 |
91
+ | Malayalam | 14,705 |
92
+
93
+ ---
94
+
95
+ # Subject Coverage
96
+
97
+ The dataset covers a broad range of educational disciplines including:
98
+
99
+ ### STEM
100
+
101
+ - Mathematics
102
+ - Physics
103
+ - Chemistry
104
+ - Biology
105
+ - Engineering
106
+ - Medical Sciences
107
+ - Computer Science
108
+ - Information Technology
109
+ - Environmental Science
110
+ - Agriculture
111
+ - Nursing
112
+ - Ayurveda
113
+ - General Science
114
+ - Quantitative Aptitude
115
+
116
+ ### Non-STEM
117
+
118
+ - History
119
+ - Geography
120
+ - Economics
121
+ - Commerce
122
+ - Business Studies
123
+ - Law
124
+ - Sociology
125
+ - Political Science
126
+ - General Knowledge
127
+ - Communication
128
+ - Teaching
129
+ - Library Science
130
+ - Management
131
+ - English Language
132
+ - Hindi Language
133
+ - Arabic Language
134
+ - Islamic Education
135
+ - Art & Design
136
+ - Home Science
137
+ - Social Science
138
+ - Miscellaneous Educational Topics
139
+
140
+ ---
141
+
142
+ # Largest Categories
143
+
144
+ | Category | Questions |
145
+ |-----------|----------:|
146
+ | Miscellaneous | 1,070,040+ |
147
+ | General Knowledge | 831,800+ |
148
+ | General English | 477,000+ |
149
+ | Reasoning | 725,000+ |
150
+ | Quantitative Aptitude | 507,000+ |
151
+ | Mathematics | 438,000+ |
152
+ | Engineering | 390,000+ |
153
+ | Medical | 161,000+ |
154
+ | Physics | 167,000+ |
155
+ | Biology | 149,000+ |
156
+ | Chemistry | 139,000+ |
157
+
158
+ ---
159
+
160
+ # Data Format
161
+
162
+ Each record contains structured educational information similar to:
163
+
164
+ ```json
165
+ {
166
+ "question": "...",
167
+ "options": [
168
+ "...",
169
+ "...",
170
+ "...",
171
+ "..."
172
+ ],
173
+ "answer": "...",
174
+ "explanation": "...",
175
+ "language": "English",
176
+ "subject": "Physics"
177
+ }
178
+ ```
179
+
180
+ ---
181
+
182
+ # Features
183
+
184
+ - 6.7M+ verified educational Q&A pairs
185
+ - Detailed explanations for every answer
186
+ - Multi-language support
187
+ - STEM and Non-STEM coverage
188
+ - Competitive examination content
189
+ - Higher education material
190
+ - School curriculum
191
+ - College curriculum
192
+ - Government exam preparation
193
+ - Mathematical equations in LaTeX
194
+ - Mathematical equations in MathML
195
+ - Embedded educational illustrations
196
+ - PDF source documents
197
+ - Structured JSON files
198
+ - Ideal for instruction tuning
199
+ - Ready for Retrieval-Augmented Generation (RAG)
200
+
201
+ ---
202
+
203
+ # Potential AI Applications
204
+
205
+ - Large Language Model (LLM) Training
206
+ - Supervised Fine-Tuning (SFT)
207
+ - Instruction Tuning
208
+ - Retrieval-Augmented Generation (RAG)
209
+ - Educational Chatbots
210
+ - AI Tutors
211
+ - Question Answering Systems
212
+ - Reasoning Models
213
+ - Benchmark Creation
214
+ - Knowledge Distillation
215
+ - Curriculum Learning
216
+ - Educational Search Engines
217
+ - Adaptive Learning Systems
218
+ - AI Exam Preparation Platforms
219
+
220
+ ---
221
+
222
+ # Industries
223
+
224
+ - Artificial Intelligence
225
+ - Education Technology (EdTech)
226
+ - E-Learning
227
+ - Universities
228
+ - Schools
229
+ - Government Education
230
+ - Competitive Exam Platforms
231
+ - Research Organizations
232
+ - Knowledge Management
233
+ - Digital Libraries
234
+ - Publishing
235
+ - Enterprise AI
236
+
237
+ ---
238
+
239
+ # File Structure
240
+
241
+ ```
242
+ Dataset/
243
+ │
244
+ ├── PDFs/
245
+ │ ├── Biology.pdf
246
+ │ ├── Mathematics.pdf
247
+ │ ├── Engineering.pdf
248
+ │ └── ...
249
+ │
250
+ ├── JSONs/
251
+ │ ├── Biology.json
252
+ │ ├── Mathematics.json
253
+ │ ├── Engineering.json
254
+ │ └── ...
255
+ ```
256
+
257
+ ---
258
+
259
+ # Dataset Statistics
260
+
261
+ - **Questions:** 6.7M+
262
+ - **English Questions:** 2.5M+
263
+ - **Languages:** 12
264
+ - **Subjects:** 45+
265
+ - **Average Question Length:** ~210 words
266
+ - **Total Tokens:** 1.8 Billion+
267
+ - **Formats:** PDF & JSON
268
+ - **Supports Images:** Yes
269
+ - **Supports Equations:** LaTeX & MathML
270
+
271
+ ---
272
+
273
+ # Advantages
274
+
275
+ - Extremely large multilingual educational corpus
276
+ - Rich reasoning-based explanations
277
+ - Suitable for both language and vision-language models
278
+ - Covers school, university, and professional education
279
+ - Balanced STEM and Non-STEM content
280
+ - Ready for scalable AI training pipelines
281
+ - Well-structured JSON annotations
282
+ - Human-readable PDF references
283
+
284
+ ---
285
+
286
+ # Limitations
287
+
288
+ - Dataset is intended primarily for educational and research applications.
289
+ - Some subjects contain significantly more samples than others.
290
+ - Coverage varies across supported languages.
291
+
292
+ ---
293
+
294
+ # License
295
+
296
+ This dataset is distributed under the **Creative Commons Attribution 4.0 International (CC BY 4.0)** License.
297
+
298
+ Users are free to use, modify, distribute, and build upon the dataset with appropriate attribution.
299
+
300
+ ---
301
+
302
+ # Citation
303
+
304
+ ```bibtex
305
+ @dataset{multilingual_qa_dataset,
306
+ title={Multilingual STEM & Non-STEM Question Answering Dataset},
307
+ year={2026},
308
+ license={CC BY 4.0}
309
+ }
310
+ ```
311
+
312
+ ---
313
+
314
+ # Conclusion
315
+
316
+ The **Multilingual STEM & Non-STEM Question Answering Dataset** provides one of the most comprehensive educational corpora currently available for AI development. With **6.7 million+ questions**, **1.8 billion+ tokens**, extensive multilingual coverage, detailed explanations, and support for mathematical notation and educational imagery, it serves as a high-quality resource for training next-generation Large Language Models, educational assistants, reasoning systems, and intelligent tutoring applications.