| --- |
| license: cc-by-4.0 |
| task_categories: |
| - table-question-answering |
| - document-question-answering |
| - feature-extraction |
| - text-generation |
| - translation |
| - summarization |
| language: |
| - id |
| - en |
| tags: |
| - finance |
| - indonesia |
| - idx |
| - stock-market |
| - financial-statements |
| - xbrl |
| - ocr |
| - multimodal-rag |
| - accounting |
| - corporate-reporting |
| --- |
| |
| # IDX_Financial_Statements: The Multimodal Indonesian Financial Dataset |
|
|
| **IDX_Financial_Statements** is a centralized repository providing the most complete set of financial disclosures for public companies listed on the **Indonesia Stock Exchange (IDX)**. This dataset is designed for advanced financial research, spanning from raw document archival to structured data extraction. |
|
|
| ## Dataset Overview |
|
|
| This is a multimodal dataset that captures the full lifecycle of financial reporting. It includes: |
|
|
| 1. **PDF Reports**: Original Annual Reports (Laporan Tahunan) and Financial Statements as submitted by issuers. Ideal for Document AI, layout analysis, and OCR tasks. |
| 2. **Excel Sheets**: Semi-structured spreadsheets containing balance sheets, income statements, and cash flow details. |
| 3. **XBRL Instances**: High-precision, machine-readable XML-based data using the IDX Taxonomy. This is the gold standard for quantitative data integrity. |
| 4. **Metadata**: Associated tickers, industry sectors, reporting periods (Q1, Q2, Q3, FY), and submission timestamps. |
|
|
| ## Key Features |
|
|
| - **Format Diversity**: Supports a wide range of technical workflows, from computer vision (PDF parsing) to direct database ingestion (XBRL). |
| - **Historical Depth**: Comprehensive coverage of the Indonesian capital market over multiple fiscal years. |
| - **Bilingual Context**: Documents often contain side-by-side Indonesian and English text, making this a valuable resource for financial-domain NMT (Neural Machine Translation). |
| - **Audit-Ready**: Includes original source documents, allowing for the verification of extracted data against official corporate filings. |
|
|
| ## Use Cases |
|
|
| - **Multimodal RAG**: Build AI agents that can "read" a 200-page PDF annual report and cross-reference it with numerical Excel data. |
| - **XBRL Analysis**: Utilize the standardized taxonomy for high-speed fundamental screening without the need for traditional scraping. |
| - **Document AI Training**: Fine-tune models for table detection, key-value pair extraction, and financial entity recognition. |
| - **Market Intelligence**: Analyze corporate governance statements and management discussion & analysis (MD&A) sections for sentiment and strategic shifts. |
|
|
| ## Data Structure |
|
|
| The repository is organized by `Ticker` and `Period`. A typical directory structure looks like: |
| - `[TICKER]/[YEAR]/[PERIOD]/` |
| - `FinancialStatement-202X-I-Ticker.pdf` |
| - `FinancialStatement-202X-I-Ticker.xlsx` |
| - `FinancialStatement-202X-I-Ticker.zip` (XBRL) |
|
|
| ## License |
|
|
| This dataset is licensed under the [Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/). You are free to share and adapt the material for any purpose, including commercial use, provided you give appropriate credit. |
|
|
| ## Disclaimer |
|
|
| This dataset aggregates publicly available information for research and analytical purposes. Users are encouraged to cross-reference data with the [IDX official disclosures](https://www.idx.co.id) for formal investment or legal decisions. |
|
|
| --- |
|
|
