Alert Text Detector 🚨📱

Project Overview

Alert Text Detector is an advanced Natural Language Processing (NLP) model designed to identify and flag emergency or critical alert messages across social media platforms. Leveraging state-of-the-art machine learning techniques, this project aims to provide real-time detection of urgent communications.

Key Features

• 🧠 Advanced NLP Model: Utilizes Bertweet Base model for sophisticated text classification
• 🔍 High Precision Detection: Achieves over 93% precision in identifying alert texts
• 📍 Location-Based Alerts: Customizable alert distribution based on user interests and locations
• 🔬 Comprehensive Preprocessing: Includes lemmatization, keyword extraction, and hashtag analysis

Performance Metrics

Validation Results

• Accuracy: 91.07%
• Precision: 89.46%
• Recall: 91.05%
• F1 Score: 90.25%

Test Dataset Results

• Accuracy: 96.37%
• Precision: 93.17%
• Recall: 99.83%
• F1 Score: 96.39%

Technology Stack

• Machine Learning: Hugging Face Transformers, Bertweet Base Model
• Data Processing: Pandas, NumPy, NLTK
• Model Training: PyTorch
• Web Framework: Flask
• Database: SQLite3

Data Gathering

This project's core dataset was built from a combination of a ready-made source and a strategically designed synthetic dataset. The goal was to train a model that could handle the nuances of real-world language, including sarcasm, historical references, and diverse alert categories, which are often sources of misclassification.

The synthetic data was meticulously created across the following categories:

Handling Ambiguity & Sarcasm: Tweets with "negative" keywords that were not actual alerts (e.g., "What a disaster!" in a movie review or sarcastic remarks).

Temporal Context: Tweets mentioning past events that should not be flagged as current alerts (e.g., "The storm last year...").

Comprehensive Alert Categories: A wide range of positive alert tweets covering various scenarios, including:

NATURAL DISASTERS

HEALTH EMERGENCIES

CIVIL AND POLITICAL UNREST

INFRASTRUCTURE FAILURES

CRIME AND SAFETY ALERTS

FINANCIAL EMERGENCIES

Varied Tweet Structures: Data was designed to mimic real social media posts, including tweets with and without hashtags, locations, and different points of view (e.g., first-person vs. third-person announcements).

This approach ensured the model was trained on a rich, diverse dataset that directly addressed common NLP challenges, leading to higher accuracy and precision. Following data collection, a comprehensive preprocessing pipeline was applied to clean and prepare the text for the model.

Installation

1. Clone the repository

git clone https://github.com/saivasanthg/Alert_Text_Detector.git
cd Alert_Text_Detector

Data Preprocessing Techniques

• Lemmatization
• Keyword extraction using paraphrase-MiniLM-L6-v2
• Hashtag analysis
• Train-test split (80% training, 20% validation)

TRAINING PROCESS

Future Roadmap

• Multi-language support
• Real-time streaming alert detection
• Integration with more social media platforms
• Decoding of emoticons for better contextual understanding

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Sai Vasanth G gsvsanth2004@gmail.com