furkancekic/turkish-stockmarket-instructions

--- language: - tr license: mit task_categories: - question-answering - text-generation tags: - finance - turkish - alpaca - instruction-tuning - capital-markets size_categories: - 10K<n<100K configs: - config_name: default data_files: - split: train path: train.json - split: eval path: eval.json --- # Turkish Stock Market Instructions Dataset Turkish financial regulation instruction-tuning dataset generated from 100+ official PDFs (Takasbank, MKK, Borsa Istanbul, SPK). ## Dataset Summary - **Train**: 66,080 instruction-output pairs - **Eval**: 7,342 instruction-output pairs - **Language**: Turkish - **Format**: Alpaca (instruction, input, output) - **Estimated Tokens**: ~9M ## Format ```json { "instruction": "Takasbank'ın teminat yönetimi nasıl işler?", "input": "", "output": "Takasbank'ın teminat yönetimi, merkezi karşı taraf hizmeti..." } ``` ## Source Generated from Turkish capital markets regulation PDFs using Gemini 2.5 Flash API. ## Intended Use Fine-tuning small language models (SLMs) for Turkish financial domain Q&A.