aayush97/semeval2023-afrisenti

A low-resource sentiment analysis project for African Languages # semeval2023-afrisenti A low-resource sentiment analysis project for African Languages ### Running the code 1. Clone the repo and `cd` to project directory 2. Install python -- version 3.8.12 2. Run the command `python -m venv .venv` to setup the virtual environment 3. Activate the virtual environment using `source .venv/bin/activate` 4. Install the requirements `pip install -r requirements.txt` 5. To run lexicon-based sentiment analysis for Igbo language, go to "Lexicon-based SA for Igbo" section below ### Training the model The following commandline options are available for training the model ``` Usage: python -m src.models.train_model [OPTIONS] Options: --lang [am|dz|ha|ig|ma|pcm|pt|sw|yo] --model [LinearSVM|NaiveBayes|naija-roberta-large|xlm-roberta-small] --finetune_lm Finetune the language model as well --finetune_classifier Finetune classification layer --help Show this message and exit. ``` The `--model` option indicates which model to train. The `--finetune_classifier` would fine tune the pretrained model on the training data. It is only applicable is the model is either `naija-roberta-large` or `xlm-roberta-small`. When `LinearSVC` or `NaiveBayes` is selected this option is ignored. The `--finetune_lm` option will finetune the masked language model objective with the traning data for that particular model. It is always used with `--finetune_classifier`. Example Usage: `python -m src.models.train_model --lang="pcm" --model="naija-roberta-large" --finetune_classifier --finetune_lm` ### Evaluating the model The following commandline options are available for evaluating the model ``` Usage: python -m src.models.predict_model [OPTIONS] Options: --lang [am|dz|ha|ig|ma|pcm|pt|sw|yo] --model [LinearSVM|NaiveBayes|naija-roberta-large|xlm-roberta-small] --finetune_classifier Use finetuned classification layer --help Show this message and exit. ``` The `--model` o …