Ever Wondered How Your Smartphone, Laptop Show Advertisement Of Things That You Need? Dr. Praveen Chand Kolli Explains

Have you ever wondered how search engines and online platforms seem to understand your preferences so well? Or how do those advertisements magically appear on your screen, tailored specifically to your interests? The answer lies in the fascinating world of deep learning, a branch of artificial intelligence that has revolutionized search and ad recommendation systems. While AI is rising very rapidly, it has been at the core of advertising for quite some years now. AI and ML together pose a combination that has been helping e-commerce players and advertisers reach a suitable audience.

Experts are of the opinion that going forward, online advertisements would get more precise and targeted and AI will have a larger role to play in that. Dr. Praveen Chand Kolli, an expert in the field who has extensive experience in building deep learning models for ad recommendations at esteemed companies, shared that platforms like Amazon or Flipkart have millions of products, and finding the most relevant item quickly is a challenging task.

“To simplify this, each item on Amazon has an "index card" equivalent containing important details like the item name, description, specifications, and manufacturer. These index cards act as pointers to help you find items efficiently, similar to subject and title cards in the library analogy. When you search for an item on Amazon, your search query is matched against different attributes of these index cards, and the system quickly sifts through the organised collection of items to retrieve roughly a thousand relevant items,” said Dr Kolli.

Similarly, social media platforms like Facebook use a similar concept of index cards to store information about advertisements. With millions of ads on their platform, Facebook needs to swiftly navigate through this vast collection to present the most relevant ads to users. “The index cards for ads are structured and categorised based on factors like interests (electronic ads, sports ads, politics ads, business ads, etc.), age (ads targeted at different age groups), and location (ads tailored for specific regions),” said Dr Kolli.

He shared that when a user visits Facebook, the platform leverages these indexed cards to efficiently filter through the large collection of ads and retrieve roughly a thousand relevant ads based on the user's interests, age, and location.

“This initial stage of the search and ad recommendation process is known as retrieval. During this step, the system effectively searches through a large collection of items or ads and retrieves approximately a thousand highly relevant results. It can be seen as a coarse-grained search. However, to ensure an optimal user experience, these platforms must sort these results based on their relevance. This is where deep learning plays a crucial role. To simplify the discussion, let's focus on the scenario of Facebook ad recommendations, but the same concept applies to Amazon item recommendations as well,” he said.

Deep learning models play a crucial role in this process by leveraging various user attributes, ad attributes, and contextual factors such as time of the day, day of the week, and location. These models take these attributes as inputs, along with user ad interaction attributes, and generate a likelihood score that indicates the probability of a specific user engaging with a particular ad.

“By considering the user's preferences and interests over different time periods, these Deep learning models generate more accurate likelihood scores using historical data. This ranking/sorting process often referred to as fine-grained search or ranking, ensures that the ads with higher scores appear at the top, while ads with lower scores are pushed down or displayed on subsequent pages,” he said.

The remarkable understanding of our preferences by search engines and online platforms, as well as the tailored advertisements we encounter, can be attributed to the advancements in deep learning. This branch of artificial intelligence has revolutionized search and ad recommendation systems, providing users with a personalized and efficient experience.