Using Siamese Networks for Advertisements Classification

Given the constant growth of investment in online advertising, it is essential to have tools that provide complete, accurate, verified, and real-time monitoring of the ads companies place on different websites and social networks to sell their products through campaigns. For this reason, many advertisement companies provide a digital platform that allows easy and secure access to strategic information for advertising agencies, media, media centers, and advertisers so that they can verify that their advertising campaigns are being carried out successfully. However, a challenge that many companies face today is the lack of semi-automatic mechanisms to identify advertising from their clients and even detect competing advertisers who invest their money in other media. Currently, this identification is accomplished by many companies manually, representing a high cost in time and effort. In this context, we propose a siamese networks approach for automatically classifying online advertisements. We evaluated the approach's performance using promotional advertising data provided by leading advertisement company in Argentina. We found that the accuracy of the proposed method can achieve an outstanding performance of 90%.