Search Results for author:
Found 5 papers, 3 papers with code
Load What You Need: Smaller Versions of Mutililingual BERT
The obtained results confirm that we can generate smaller models that keep comparable results, while reducing up to 45% of the total number of parameters.
Attention over pre-trained Sentence Embeddings for Long Document Classification
When compared with the current state-of-the-art models using standard fine-tuning, the studied method obtains competitive results (even if there is no clear best model in this configuration).
DziriBERT: a Pre-trained Language Model for the Algerian Dialect
The obtained results show that pre-training a dedicated model on a small dataset (150 MB) can outperform existing models that have been trained on much more data (hundreds of GB).
Load What You Need: Smaller Versions of Multilingual BERT
The obtained results confirm that we can generate smaller models that keep comparable results, while reducing up to 45% of the total number of parameters.
Classification de relations pour l'intelligence \'economique et concurrentielle (Relation Classification for Competitive and Economic Intelligence )
L{'}{\'e}valuation de ces mod{\`e}les montre des r{\'e}sultats tr{\`e}s encourageants, ce qui est un premier pas vers l{'}intelligence {\'e}conomique et concurrentielle {\`a} partir de textes pour le fran{\c{c}}ais.
