1. Abstract

In the digital marketing industry, video advertisements play an important role in promoting products and services. Every day, a large number of video ads are produced in different categories such as travel, food, entertainment, and lifestyle. Manually identifying and classifying these advertisements is time-consuming and inefficient.

This project focuses on developing a deep learning model to classify video advertisements based on their textual transcriptions. The text data is extracted from video ads and preprocessed for training. An LSTM (Long Short-Term Memory) neural network is used to analyze the sequential nature of text data and predict the category of the advertisement.

The trained model is integrated with a Django web application and deployed on Google Cloud. This allows users to upload or enter ad text and receive instant category predictions.

2. Objectives

The main objectives of this project are:

1. To understand text-based classification techniques.
2. To study deep learning concepts using LSTM.
3. To preprocess and clean text data.
4. To convert text into numerical format.
5. To train a deep learning classification model.
6. To predict advertisement categories.
7. To develop a web-based interface.
8. To integrate the model with Django.
9. To deploy the application on cloud platform.
10. To improve knowledge in NLP and deep learning.

3. Existing System

In the existing system, video advertisements are classified manually or using simple rule-based methods.

The limitations of the existing system are:

1. Manual classification is time-consuming.
2. High chances of human error.
3. Not suitable for large datasets.
4. Low accuracy.
5. No automatic learning.
6. Poor handling of complex text.
7. Limited scalability.

These drawbacks reduce efficiency and increase operational cost.

4. Proposed System

The proposed system uses deep learning and NLP techniques to classify video advertisements automatically.

In this system:

• Advertisement transcriptions are collected.

• Text data is cleaned and preprocessed.

• Data is converted into numeric format.

• LSTM model is trained.

• Categories are predicted.

• Django web interface is created.

• Application is hosted on Google Cloud.

This system provides accurate and fast advertisement classification.

5. Implementation Procedure

The project is implemented using the following steps:

Step 1: Data Collection

Collect video advertisement transcriptions from datasets.

Step 2: Data Loading

Load the dataset into Python environment.

Step 3: Data Preprocessing

• Remove stop words

• Convert to lowercase

• Remove special characters

• Tokenization

• Padding sequences

Step 4: Feature Extraction

Convert text into numerical form using embedding techniques.

Step 5: Model Building

Build LSTM neural network model.

Step 6: Model Training

Train the model using labeled data.

Step 7: Web Application Development

Develop Django-based user interface.

Step 8: Deployment

Deploy application using Google App Engine.

6. Software Requirements

The software tools required for this project are:

• Python

• Jupyter Notebook / VS Code

• TensorFlow / Keras

• NumPy, Pandas

• NLTK / SpaCy

• Django Framework

• Google Cloud Platform

• Web Browser

7. Hardware Requirements

The hardware requirements include:

• Processor: Intel i5 or higher

• RAM: 8 GB or higher

• Storage: 256 GB or higher

• System: Laptop/Desktop

• Internet Connection: Stable broadband

Optional:

• GPU for faster training

8. Advantages of the Project

1. Automates advertisement classification.
2. Reduces manual effort.
3. Provides high accuracy.
4. Works with large datasets.
5. Supports real-time prediction.
6. Easy-to-use web interface.
7. Cloud accessible system.
8. Enhances deep learning skills.
9. Useful for marketing analytics.
10. Improves business decision-making.