| | --- |
| | license: cc-by-sa-3.0 |
| | --- |
| | ## Semantic search over the 44 million of English Wikipedia paragraphs using sentence transformers encoder. |
| |
|
| | The dataset contains: |
| | - 43 911 155 paragraphs from 6 458 670 wikipedia articles stored in a zip archive; |
| | - FAISS index with the embeddings; |
| | - Retriever module for semantic search over the paragraphs. |
| |
|
| | The size of each paragraph varies from 20 to 2000 characters. |
| | The embedding vector size is 768. |
| | The index is 4-bit-quantized 2-level IVF16384_HNSW32 constructed with the [FAISS library](https://github.com/facebookresearch/faiss). |
| | Sentence encoder: [all-mpnet-base-v2](https://huggingface.co/sentence-transformers/all-mpnet-base-v2). |
