rasyosef/amharic-sentiment-classification

notebooks to finetune Amharic BERT and RoBERTa models for sentiment classification ### Amharic Sentiment Classification The following models were finetuned using the amharic-sentiment dataset. The *finetuning notebooks* can be found in the `notebooks` folder. The reported precision, recall, and f1 metrics are macro averages. |Model|Size (# params)| Accuracy | Precision | Recall | F1 | | --- | ------------- | -------- | --------- | ------ | -- | |roberta-base-amharic|110M|**0.88**|**0.88**|**0.88**|**0.88**| |roberta-medium-amharic|42.2M|0.84|0.84|0.84|0.84| |bert-medium-amharic|40.5M|0.83|0.83|0.82|0.83| |bert-small-amharic|27.8M|0.83|0.83|0.82|0.83| |bert-mini-amharic|10.7M|0.81|0.81|0.81|0.81| |bert-tiny-amharic|4.18M|0.79|0.79|0.79|0.79| |xlm-roberta-base|279M|0.83|0.83|0.83|0.83| |afro-xlmr-base|278M|0.83|0.83|0.83|0.83| |afro-xlmr-large|560M|0.86|0.86|0.86|0.86| |am-roberta|443M|0.82|0.83|0.82|0.82| ### Fine-tuned Model A finetuned `bert-medium-amharic` model is available on HuggingFace: rasyosef/bert-medium-amharic-finetuned-sentiment #### How to use You can use the model directly with a pipeline for text classification: ```python >>> from transformers import pipeline >>> bert_sentiment = pipeline("text-classification", model="rasyosef/bert-medium-amharic-finetuned-sentiment") >>> bert_sentiment(["አሪፍ ፊልም ነው።", "ዩክሬን እና ሩስያ ከባድ ውግያ ላይ ናቸው።"]) [{'label': 'positive', 'score': 0.9863048791885376}, {'label': 'negative', 'score': 0.9570127129554749}] ``` ### References Fine-tuning a model with the Trainer API