This project demonstrates how to perform Named Entity Recognition (NER) using a Conditional Random Field (CRF) model. It includes steps for preprocessing data, feature extraction, model training, evaluation, and predictions.

• Dataset Preparation: Loads and preprocesses data into BIO-tagged format suitable for NER tasks.
• Feature Engineering: Extracts token-level features such as:
  • Part-of-speech tags
  • Word casing and suffixes
  • Previous and next word context
• Model Training: Uses sklearn_crfsuite to train a CRF model.
• Evaluation: Computes precision, recall, F1-score, and supports detailed performance analysis.
• Prediction: Allows for inference on new sequences using the trained model.

1. Clone the repository or download the notebook.
2. Make sure the required libraries are installed.
3. Run main-NER-CRF.ipynb to execute the full pipeline.

After training, the notebook prints a detailed classification report, such as:

precision    recall  f1-score   support

     B-PER       0.95      0.93      0.94       200
     I-PER       0.90      0.92      0.91       150
     ...

• You can easily modify the feature function to include domain-specific features.
• The notebook is modular, allowing you to plug in other sequence labeling algorithms (e.g., BiLSTM-CRF).

• main-NER-CRF.ipynb - Main notebook with all steps from data preprocessing to evaluation.
• person_names.txt - List of common person names.
• locations.txt - List of cities and countries.