DjebbiAmine/Senti-BERT

A pre-trained BERT model for multi-dialect Sentiment Analysis of North African dialects This is a repository of the Multi-dialect Sentiment Analysis BERT model. By AmineDjebbi. ## About our Multi-dialect-Sentiment-BERT model: This repository represents the first MultiDialect Attentive model Sentiment Analysis from BERT tuned from scratch and trained on 80k tweets of four different datasets (Tunisian, Egyptian, Algerian and Levant). ## Usage: The model weights can be loaded using transformers library by HuggingFace. model = AutoModel.from_pretrained("aminedjebbie/SentiBERT") ## Model Parameters: | Parameter | Value | | ------------- | ------------- | | architecture | Uncased_L8_H12| | hidden_size | 768 | | max_position_embeddings | 512 | | num_attention_heads | 8 | | num_hidden_layers | 12 | | vocab_size | 51647 | | Total number of parameters | 110M |