| # Netflix Technical Documentation Dataset | |
| ## Overview | |
| This dataset consists of high-quality PDF page images from technical Netflix documentation, paired with technical queries in multiple languages (French, English, Spanish, German, and Italian). It's designed for training and evaluating multimodal retrieval models that can understand both textual queries and visual document content. | |
| ## Dataset Structure | |
| The dataset is divided into two main components: | |
| - **test-comp-aer.parquet**: Contains query-document pairs with the following fields: | |
| - `query`: Technical query about Netflix operations/technology | |
| - `ranked_documents`: JSON string containing document identifiers | |
| - `image`: High-resolution image of the relevant PDF page | |
| - **train-comp-aer.parquet**: Contains document images with the following fields: | |
| - `file_name`: PDF source filename | |
| - `page`: Page number in the source document | |
| - `query`: Associated technical query | |
| - `image`: High-resolution image of the PDF page | |
| ## Use Cases | |
| This dataset is particularly useful for: | |
| - Training multimodal retrieval systems that can search technical documentation | |
| - Developing document understanding models that combine text and visual elements | |
| - Building multilingual technical document search systems | |
| - Researching document question-answering with visual context | |
| ## Creation Methodology | |
| This dataset was created using a specialized pipeline that: | |
| 1. Processes technical PDF documents from Netflix | |
| 2. Extracts high-resolution images of PDF pages | |
| 3. Generates diverse technical queries using specialized large language models | |
| 4. Maintains multilingual query distribution across French, English, Spanish, German, and Italian | |
| 5. Assembles the data into structured parquet files for efficient processing | |
| ## About Me | |
| I am Khatir SEDJAI, a third-year undergraduate student passionate about artificial intelligence. I enjoy creating datasets and models that push the boundaries of what AI can accomplish. | |
| My interests include multimodal learning, document understanding, and building practical AI applications. This dataset represents my contribution to advancing multimodal document retrieval capabilities. | |
| I'm particularly focused on bridging the gap between textual queries and visual document understanding, with specific attention to technical documentation. | |
| Connect with me on [LinkedIn](https://www.linkedin.com/in/khatir-sedjai-246766242) to discuss this dataset or potential collaborations! | |
| ## License | |
| [Add appropriate license information here] | |
| ## Citation | |
| If you use this dataset in your research, please consider citing: | |
| ``` | |
| @dataset{sedjai2025netflixdataset, | |
| author = {Sedjai, Khatir}, | |
| title = {Netflix Technical Documentation Dataset}, | |
| year = {2025}, | |
| publisher = {Hugging Face}, | |
| howpublished = {\url{https://huggingface.co/datasets/khatir/netflix-technical-documentation}} | |
| } | |
| ``` | |
| ## Acknowledgements | |
| I would like to thank [Include any acknowledgements here - individuals, organizations, or funding sources that supported this work] |