Use cases: Gen AI, Personalization
Industries: Telecommunications, Media
Products and Tools: MongoDB Atlas, MongoDB Atlas Vector Search, MongoDB PyMongo Driver
Solution Overview
In today's evolving media landscape, the large amount of digital content makes it difficult to capture audience attention. At the same time, referral traffic from social media platforms is declining, which puts additional pressure on traditional media outlets to drive engagement. As a result, publishers are seeking ways to stabilize their user base and enhance content engagement.
To overcome these challenges, publishers need to use data effectively to create more engaging and personalized experiences for their users. With MongoDB Atlas and Atlas Vector Search, you can build an AI-driven media platform that transforms content delivery to users of large-scale media and publishing platforms. By analyzing interactions and consumption patterns, this solution grasps what content resonates with users and predicts what they are likely to engage with in the future. These insights enable publishers to construct a personalized content journey.
Reference Architectures
The architecture below shows how you can build an AI-driven media solution with MongoDB that incorporates advanced personalization services, such as:
Discovery and content suggestions
Content summarization and reformating
Keyword extraction
Insights and dossiers automation
Figure 1. AI-driven media architecture with Atlas Vector Search
The following sections describe these services in more detail.
Discovery and Content Suggestions
This solution suggests content based on users' preferences and past interactions by utilizing user data, behavioral analytics, and vectorization of media items. This enhances user engagement and increases the likelihood of converting free users into paying subscribers. MongoDB's Vector Search performs kNN searches, optimizing how content is matched by embedding vectors directly into MongoDB documents. This means that you don't need to manage multiple applications or transfer data between different database systems, which simplifies the architecture. Additionally, MongoDB's scalability and resilience enable organizations to scale their operations vertically or horizontally. You can also scale search nodes independently from operational database nodes to adapt to the specific load scenario.
Content Summarization and Reformating
Users have a diverse array of consumption habits. This solution provides concise summaries and adapts content formatting based on user preferences and device specifications.
Keyword Extraction
In traditional publishing workflows, selecting keywords requires content creators to meticulously identify and incorporate relevant keywords. This process can be time-consuming and prone to human error, as significant keywords can be overlooked which can diminish the content's visibility and engagement.
With the help of the underlying LLM, this solution draws essential information through keyword extraction, enabling users to grasp key news dimensions and enhancing the searchability of content within the platform. Keywords significantly influence the SEO performance of digital content.
Insights and Dossiers Automation
This solution automatically generates comprehensive insights and dossiers from multiple articles, which is helpful for users who want to learn in-depth about specific topics or events. This capability uses one or more LLMs to generate natural language output, derived from multiple source articles. You can integrate any leading language model that fits your requirements. Here's how this process works:
Integration with multiple sources: The system uses Atlas Vector Search to pull content from a variety of articles and data sources. This content is then compiled into dossiers, which provide users with a detailed and contextual exploration of topics and are curated to offer a narrative or analytical perspective beyond the original content.
Customizable output: You can customize the system output by setting parameters based on your audience's preferences or your specific project requirements. This includes adjusting the level of detail, the amount of technical terms, and the inclusion of multimedia elements.
You can reuse the core concepts of this solution across other industries such as retail, where presenting the right products to the right users is essential to keep sales high.
Build the Solution
You can view the solution demo at https://ist.media, or replicate
it using the README of this GitHub repository.
Figure 2. Media platform homepage interface
In the underlying data model, a representative news article has the following structure:
Figure 3. Data model for a news article
You can use Voyage AI to generate your embeddings. To perform vector search, create a vector index in MongoDB Atlas with the following configuration:
Figure 4. Vector index for the embedding model
Key Learnings
Build AI-driven applications: With MongoDB Atlas, you can build an AI-powered media solution that delivers tailored content to your users and automate backend processes, such as keywords automation.
Store diverse data types: With MongoDB's flexible document model, you can store a wide range of media data including user data, news articles, and embeddings, which simplifies the development of AI-driven applications.
Personalize user experiences: With MongoDB Atlas Vector Search, you can create a personalized content journey, based on individual preferences and past interactions, that improves user engagement.
Authors
Benjamin Lorenz, MongoDB
Learn More
To learn how to build smarter searches, visit the Atlas Vector Search Quick Start guide.