krish5932/Customer-FeedBack-System

Smart Customer Feedback & Topic Clustering Dashboard

A premium, interactive Natural Language Processing (NLP) dashboard built with Streamlit to analyze customer reviews, classify sentiment, group reviews into unsupervised topic clusters, and evaluate machine learning models.

🚀 Key Features

1. Advanced NLP text Preprocessing

• —Contraction Expansion: Maps contractions (e.g. "don't" $\rightarrow$ "do not", "can't" $\rightarrow$ "cannot") before cleaning.
• —Negation Preservation: Preserves sentiment-critical negation terms (e.g., "not", "no", "never", "nor") from stopword stripping, ensuring phrases like "not good" are classified correctly.
• —Lemmatization: Standardizes words to their dictionary forms using NLTK's WordNetLemmatizer.

2. Multi-Model Sentiment Classification

• —Choose between three popular classifiers from Scikit-Learn:
• —Logistic Regression (Balanced class weights)
• —Linear Support Vector Classifier (LinearSVC) (Calibrated for probability estimation)
• —Multinomial Naive Bayes
• —Tune hyperparameters (Regularization strength $C$, Smoothing $\alpha$) directly from the sidebar.

3. Model Performance & Diagnostics

• —Evaluates classifiers on an 80/20 train-test split to report realistic, unbiased out-of-sample accuracy.
• —Generates a Seaborn Confusion Matrix Heatmap showing true vs. predicted sentiment frequencies.
• —Renders a detailed Classification Report table outlining Precision, Recall, and F1-score for positive, negative, and neutral classes.

4. Topic Clustering & Silhouette Optimization

• —Vectorizes feedback using TF-IDF (unigrams & bigrams).
• —Uses K-Means Clustering to discover hidden topic groups.
• —Projects the high-dimensional document vectors into a 2D space using PCA (Principal Component Analysis) for spatial visualization.
• —Cluster Count Optimizer: Runs an on-demand Silhouette Score Analysis for $K \in [2, 8]$ and graphs the score. The peak value suggests the mathematically optimal number of topics for the active dataset.

5. Live NLP Pipeline Tester

• —Run the entire NLP pipeline interactively.
• —Input custom reviews and instantly view:
• —Preprocessed tokens side-by-side with original text.
• —Sentiment predictions and confidence breakdown.
• —Topic assignment and keywords.
• —Visual projection of the entry (marked by a star) in 2D semantic space.

🛠️ Installation & Setup

Prerequisites

• —Python 3.8 or higher

Steps

1. 1.Clone the repository:

   git clone https://github.com/krishjuit/Customer-Feedback-System.git
   cd Customer-Feedback-System

1. 1.Set up a virtual environment (Recommended):

   python -m venv .venv
   # Activate on Windows:
   .venv\Scripts\activate
   # Activate on macOS/Linux:
   source .venv/bin/activate

1. 1.Install the dependencies:

   pip install -r requirements.txt

1. 1.Launch the Streamlit app:

   streamlit run app.py

📁 Project Structure

• —app.py: Streamlit frontend dashboard layout, UI blocks, and plotting logic.
• —model_utils.py: Text cleaning preprocessing functions, SentimentModel pipelines, and TopicClustering algorithms.
• —data_generator.py: Generates the default synthetic customer review dataset.
• —feedback_dataset.csv: The default loaded customer dataset containing 4000+ customer reviews.
• —requirements.txt: List of dependencies.
• —.gitignore: List of untracked file patterns to ignore.