Text-Classification-using-TensorFlow

Wine Reviews Sentiment Analysis

This project aims to classify wine reviews based on their descriptions as either high-rated (90 points or above) or low-rated (below 90 points) using Natural Language Processing (NLP) techniques and neural networks. The project utilizes TensorFlow and TensorFlow Hub to create a text classification model.

Dataset

The dataset used for this project is a CSV file named wine-reviews.csv, which contains wine reviews along with associated metadata. The relevant columns used in this project are:

• country: The country of origin of the wine.
• description: The textual description of the wine.
• points: The rating of the wine on a scale of 1 to 100.
• price: The price of the wine.
• variety: The variety of the wine.
• winery: The winery that produced the wine.

Preprocessing

1. Loading the Data: The dataset is loaded, and rows with missing values in the description and points columns are removed.

2. Creating Labels: A new binary label is created where a wine is labeled as 1 if it has a rating of 90 points or above and 0 otherwise.

3. Splitting the Dataset: The dataset is randomly split into three parts:

   • Training Set: 80% of the data
   • Validation Set: 10% of the data
   • Test Set: 10% of the data

Modeling

Two models are developed:

1. Model 1: Pretrained Embedding Layer

   • A pre-trained word embedding layer from TensorFlow Hub (nnlm-en-dim50/2) is used.
   • The model consists of:
     • A dense layer with ReLU activation
     • A dropout layer with a rate of 0.4
     • Another dense layer with ReLU activation
     • A dropout layer with a rate of 0.4
     • A final dense layer with a sigmoid activation for binary classification
   • Optimization: The model is compiled using the Adam optimizer, with binary cross-entropy as the loss function and accuracy as the evaluation metric.

2. Model 2: Custom Embedding with LSTM

   • A custom text vectorization layer is used to tokenize and vectorize the text data.
   • The model consists of:
     • An embedding layer
     • An LSTM layer with 32 units
     • A dense layer with ReLU activation
     • A dropout layer with a rate of 0.4
     • A final dense layer with a sigmoid activation for binary classification
   • Optimization: Similar to Model 1, this model uses the Adam optimizer with binary cross-entropy as the loss function and accuracy as the evaluation metric.

Training

Each model is trained for 5 epochs, and the performance is evaluated on the validation set. After training, the model is evaluated on the test set to determine its generalization performance.

Usage

To run the project, follow these steps:

1. Install Dependencies: Ensure you have Python installed along with the necessary libraries.

   pip install numpy pandas matplotlib tensorflow tensorflow_hub

2. Run the Script: Execute the script in a Python environment to train the models and evaluate their performance.

3. Visualize the Data: A histogram of the wine points is plotted for initial exploratory analysis.

Conclusion

This project demonstrates how to build and evaluate text classification models using pre-trained embeddings and custom embeddings with LSTM in TensorFlow. The performance of these models can be further improved with hyperparameter tuning, additional data preprocessing, or more advanced architectures.

---

Author: Sahil Awatramani

---