Architecture for detecting advertisement types

Abstract

 Businesses often encounter difficulties in accurately categorizing advertisements based on their content, which leads to inefficiencies in targeted marketing and data analysis. An automated system is needed to detect the type of advertisement (e.g., food, beverage, clothing, etc.) from the submitted text. This paper aims to develop an automated system leveraging Amazon Web Services (AWS) to categorize advertisement types based on text input. We employed Convolutional Neural Networks (CNNs) in developing the automated system due to the significant performance of CNNs leveraging AWS. The solution aims to boost marketing efficiency and strengthen data analysis capabilities. To achieve this, we utilize Amazon Simple Storage Service (S3), AWS Lambda, Amazon Comprehend, Amazon DynamoDB, Amazon CloudWatch, Amazon CloudFront, and AWS Web Application Firewall (AWS WAF). Moreover, we follow some procedural steps in executing the task by uploading the advertisement text to an S3 bucket, which triggers a Lambda function that forwards the text to Amazon Comprehend for analysis, and the results are stored in DynamoDB, from where the results notification is sent to the user. Magazine image datasets were employed as test datasets for this approach. This work enables automatic advertisement categorization, enhanced marketing effectiveness, better data analysis and reporting functionality, an affordable solution using AWS services, and instant feedback for users. The AWS-based architecture provides a dependable solution for the automatic identification of advertisement types. By leveraging various AWS services, the system ensures efficiency, precision, and scalability, ultimately enhancing marketing strategies and data management.