| --- |
| license: mit |
| language: |
| - en |
| metrics: |
| - accuracy |
| - precision |
| - recall |
| - f1 |
| tags: |
| - code |
| --- |
| |
| # Model Card for Sentiment Analysis on Primate Dataset |
|
|
| This model card provides details about a sentiment analysis model trained on a dataset containing posts related to primates. The model predicts sentiment labels for textual data using transformer-based architectures. |
|
|
| ## Model Details |
|
|
| ### Model Description |
|
|
| The sentiment analysis model aims to classify text data into sentiment categories such as positive, negative, or neutral. It utilizes transformer-based architectures for sequence classification. |
|
|
| - **Developed by:** Jaskaran Singh |
| - **Model type:** Transformer-based sentiment analysis model |
| - **Language(s) (NLP):** English |
| - **License:** MIT |
| - **Finetuned from model:** Transformer-based pre-trained model |
|
|
| ### Model Sources |
|
|
| - **Repository:** https://github.com/JaskaranSingh-01/Sentiment_Analyzer |
| - **Demo:** https://sentimentanalyzer-f76oxwautwypxpea4lj3wg.streamlit.app/ |
| |
| ## Uses |
| |
| ### Direct Use |
| |
| The model can be directly used for sentiment analysis tasks, particularly on textual data related to primates. |
| |
| ### Downstream Use |
| |
| The model can be fine-tuned for specific downstream tasks or integrated into larger applications requiring sentiment analysis functionality. |
| |
| ## Bias, Risks, and Limitations |
| |
| ### Bias |
| |
| The model's predictions may reflect biases present in the training data, including any biases related to primates or sentiment labeling. |
| |
| ### Risks |
| |
| - Misclassification: The model may misclassify sentiment due to ambiguity or complexity in the text. |
| - Generalization: The model's performance may vary across different domains or datasets. |
| |
| ### Limitations |
| |
| - Limited Domain: The model's effectiveness may be limited to text related to primates. |
| - Cultural Bias: The model's performance may be influenced by cultural nuances present in the training data. |
| |
| ## Recommendations |
| |
| Users should be cautious when interpreting the model's predictions, considering potential biases and limitations. Fine-tuning on domain-specific data or applying post-processing techniques may help mitigate biases and improve performance. |
| |
| ## How to Get Started with the Model |
| |
| ```python |
| # Example code for using the sentiment analysis model |
| |
| # 1. Load the model and tokenizer |
| from transformers import AutoTokenizer, AutoModelForSequenceClassification |
| |
| tokenizer = AutoTokenizer.from_pretrained("sbcBI/sentiment_analysis_model") |
| model = AutoModelForSequenceClassification.from_pretrained("sbcBI/sentiment_analysis_model") |
| |
| # 2. Tokenize input text |
| text = "Sample text for sentiment analysis" |
| encoded_input = tokenizer(text, return_tensors='pt') |
| |
| # 3. Perform inference |
| output = model(**encoded_input) |
| predicted_label = output.logits.argmax().item() |
| |
| # 4. Interpret prediction |
| sentiment_labels = ['Negative', 'Neutral', 'Positive'] |
| print("Predicted Sentiment:", sentiment_labels[predicted_label]) |
| ``` |
| |
| ## Training Details |
| |
| ### Training Data |
| |
| The training data consists of posts related to primates, annotated with sentiment labels. |
| |
| ### Training Procedure |
| |
| #### Preprocessing |
| |
| Text data underwent preprocessing steps including lowercase conversion, punctuation removal, tokenization, stopword removal, and stemming. |
| |
| #### Training Hyperparameters |
| |
| - **Training regime:** Fine-tuning of transformer-based pre-trained model |
| - **Optimizer:** Adam optimizer |
| - **Learning rate:** 5e-5 |
| - **Batch size:** 8 |
| - **Epochs:** 10 |
| |
| ### Evaluation |
| |
| #### Testing Data, Factors & Metrics |
| |
| - **Testing Data:** Holdout test set |
| - **Metrics:** Accuracy, Precision, Recall, F1-score |
| |
| #### Results |
| |
| - **Accuracy:** 0.79 |
| - **Precision:** 0.74 |
| - **Recall:** 0.77 |
| - **F1-score:** 0.75 |
| |
| ## Environmental Impact |
| |
| Carbon emissions were not directly measured for model training. However, users should consider the environmental impact of training and deploying machine learning models, especially on large-scale infrastructure. |
| |
| ## Technical Specifications |
| |
| ### Model Architecture and Objective |
| |
| The model architecture is based on transformer-based architectures, specifically designed for sequence classification tasks such as sentiment analysis. |
| |
| ### Compute Infrastructure |
| |
| #### Software |
| |
| - **Framework:** PyTorch |
| - **Dependencies:** Transformers, NLTK |
| |
