How omar-basheer/pii-identification-model works
Main Components
• Data Input: Represent the raw student essays as the primary data source. This could be a dataset imported from a CSV, database, or directly from Kaggle.
• Data Preprocessing: A component for text cleaning, tokenization, and formatting data for model consumption.
• Model Training & Evaluation: This is where your Bi-LSTM model is defined and trained. Although the provided file tree only shows a notebook (PII_Milestone3.ipynb), assume that within it are key processes like model definition, training routines, and evaluation metrics.
• Model Inference: Even if not explicitly separated in your current files, include a component that represents how the model can be used to identify PII in new essays.
• Experimentation/Visualization: Given that you’re working in a Jupyter Notebook, a component showing exploratory data analysis or experiment visualizations (metrics, loss curves, etc.) is helpful.
Out Relationships and Interactions
• From Data Preprocessing, show the flow into the Model Training & Evaluation component.
• Finally, link Model Inference to the end-user or a downstream application where predictions (PII identified) are returned.
Architectural Patterns and Design Principles
• Iterative Development: Indicate loops for hyperparameter tuning and model evaluation.
• Modularization: Show that each component (data handling, preprocessing, model building) is designed to stand alone, which can facilitate future extensions (like deploying the model as a microservice).
Relevant Technologies and Libraries
• Mention that the Bi-LSTM model might be implemented using popular ML frameworks such as TensorFlow/Keras or PyTorch (if applicable).
• Data preprocessing might utilise libraries such as Numpy, Pandas, or text-specific libraries (e.g., NLTK or spaCy).
• If experiment tracking or visualization is involved, reference tools like matplotlib or seaborn.