📊 AI Sentiment Engine & Data Insights Dashboard

An end-to-end Data Science application that bridges Natural Language Processing (NLP) with real-time descriptive analytics. This platform extracts semantic sentiment from unstructured text data across various formats and dynamically aggregates it into an operational intelligence dashboard.

🚀 Live Demo

[LINK]

💡 Dual-Focus Portfolio Value

This project is explicitly engineered to demonstrate competencies required across both Data Analyst and Machine Learning Engineer roles:

1. The Engineering Aspect (AI/NLP)

• Algorithmic Context Awareness: Utilises the VADER (Valence Aware Dictionary and sEntimer Reasoner) lexicon, which natively maps semantic shifts (e.g., negations like "not good", intensifiers, and punctuation).
• Deterministic Heuristics: Implements a custom keyword-weighting algorithm to address edge cases in the global news context, preventing geopolitical crises or grim articles from falsely flagging as positive due to administrative vocabulary (e.g., "objectives achieved").
• Unstructured Data Pipeline: Automates web-scraping and data ingestion from HTML URLs via Newspaper3k, text tokenisation with NLTK, and file parsing for .pdf and .docx blobs.

2. The Analytical Aspect (Data & Insights)

• Session Aggregation Lifecycle: Tracks transactional user interactions seamlessly using state management (st.session_state) to cache sequential analyses dynamically.
• Descriptive Statistics: Generates instantaneous aggregate KPIs, processing mathematical transformations like running means, mode distribution, and volumetric counts.
• Exploratory Data Analysis (EDA): Populates dynamic visualisation layers (categorical frequency bar charts and score variance tracking) to monitor semantic volatility.
• Data Portability: Features an on-demand data exporter that structures raw session logs into standardised, cleaned .csv datasets for downstream business intelligence workflows.

---

🛠️ System Architecture & Workflow

1. Ingestion Layer: Accepts unstructured inputs via active web URLs, document file uploads, or manual string entry.
2. Processing Layer: Cleans text, runs keyword extraction tokenisers, and targets context triggers.
3. Sentiment Engine: Evaluates compound polarity score ranges $[-1, 1]$.
4. Aggregation Layer: Updates a structured data ledger in memory via a Pandas backend.
5. Visualisation UI: Triggers real-time graphical renders and provides data download options.

---

📦 Project Structure

├── app.py              # Main application logic, UI design, and NLP pipelines
├── requirements.txt    # Python package dependencies (Streamlit, VADER, Pandas, etc.)
├── packages.txt        # System-level Linux dependencies for production parsing
└── nltk.txt            # Pre-downloaded NLTK tokeniser models (punkt, punkt_tab)

🛠️ Installation & Setup

1. Clone the repository:

   git clone [https://github.com/Sujata005/Sentimental-Analysis.git](https://github.com/Sujata005/Sentimental-Analysis.git)
   cd Sentimental-Analysis

2. Install Dependencies: pip install -r requirements.txt
3. Run the application: streamlit run app.py

👤 Author

Sujata Bijalwan