This is a test model that was fine-tuned using the Spanish datasets from stsb_multi_mt in order to understand and benchmark STS models.
This model was built taking
distilbert-base-uncased and training it on a Semantic Textual Similarity task using a modified version of the training script for STS from Sentece Transformers (the modified script is included in the repo). It was trained using the Spanish datasets from stsb_multi_mt which are the STSBenchmark datasets automatically translated to other languages using deepl.com. Refer to the dataset repository for more details.
This model was built just as a proof-of-concept on STS fine-tuning using Spanish data and no specific use other than getting a sense on how this training works.
You may use it as any other STS trained model to extract sentence embeddings. Check Sentence Transformers documentation.
Use the included script to train in Spanish the base model. You can also try to train another model passing it's reference as first argument. You can also train in some other language of those included in the training dataset.
distilbert-base-uncased on the Spanish test dataset before training results in:
Cosine-Similarity : Pearson: 0.2980 Spearman: 0.4008
While the fine-tuned version with the defaults of the training script and the Spanish training dataset results in:
Cosine-Similarity : Pearson: 0.7451 Spearman: 0.7364
In our STS Evaluation repository we compare the performance of this model with other models from Sentence Transformers and Tensorflow Hub using the standard STSBenchmark and the 2017 STSBenchmark Task 3 for Spanish.
- Training dataset stsb_multi_mt
- Sentence Transformers Semantic Textual Similarity
- Check sts_eval for a comparison with Tensorflow and Sentence-Transformers models
- Check the development environment to run the scripts and evaluation
- Downloads last month