e5-base-sts-en-de
| Property | Value |
|---|---|
| Parameter Count | 278M |
| License | MIT |
| Tensor Type | F32 |
| Best Performance | 0.904 Spearman correlation |
What is e5-base-sts-en-de?
e5-base-sts-en-de is a specialized language model fine-tuned for semantic textual similarity tasks, particularly focused on German language processing. Based on the multilingual-e5-base architecture, this model has been specifically optimized to understand and compare the meaning of text passages in German.
Implementation Details
The model implements a two-stage training approach: first utilizing Multiple Negatives Ranking Loss on paraphrase datasets, followed by Cosine Similarity Loss training on semantic textual similarity datasets. It's built on XLM-RoBERTa architecture and fine-tuned using three key datasets: German paraphrase corpus, PAWS-X, and STSB-Multi-MT.
- 278M parameters for robust language understanding
- F32 tensor type for precise computations
- Achieves 0.920 on STSB validation subset
- 0.904 Spearman correlation on test data
Core Capabilities
- Semantic similarity assessment for German text
- Cross-lingual text comparison
- Paraphrase detection and evaluation
- Feature extraction for NLP tasks
Frequently Asked Questions
Q: What makes this model unique?
The model's distinctive feature is its specialized fine-tuning for German semantic similarity tasks, combining multiple training datasets and achieving state-of-the-art performance on the STSB benchmark.
Q: What are the recommended use cases?
This model is ideal for applications requiring semantic similarity comparison in German text, including document similarity analysis, paraphrase detection, and cross-lingual text matching between German and English content.
