AI-Powered YouTube Video Recommender for the Austrian Chamber of Commerce (WKO Inhouse)
OpenAI
Recommendation System
Chatbots
Data Science
Gov
Client
The Austrian Chamber of Commerce (WKO Inhouse) represents over 540,000 member companies, advocating for Austrian businesses and supporting policies like tax relief, reducing red tape, and offering subsidies. WKO Inhouse provides expert advice on topics such as labor laws and customs, and strengthens the economy through its educational programs.
Review
Postdata's efforts have resulted in completed tasks of good quality and increased clicks on the client's website using modern AI approaches. The team has provided friendly communication and a flexible approach throughout the engagement. Overall, the client is satisfied with Postdata's performance. - Kateryna Salii, WKO Inhouse
Summary
In the era of abundant online content, personalized recommendations have become critical for engaging users and improving content discoverability. The Austrian Chamber of Commerce (WKO) aimed to enhance its YouTube channel by creating a personalized recommendation system that connects users with relevant content. To achieve this, it was first necessary to organize videos into clear categories and subcategories, forming the foundation for tailored recommendations. This case study highlights how Postdata leveraged OpenAI's advanced natural language processing tools to develop a personalized video recommendation system, enabling WKO to connect users with relevant content tailored to their preferences.
Project overview
Our team was approached to develop an AI-powered personalized recommendation system for the Austrian Chamber of Commerce’s YouTube channel. We built a Python-based service that utilized OpenAI’s API to analyze video metadata such as descriptions, tags, and categories. This system was designed to automate video organization and enable seamless recommendations, providing users with a more intuitive and personalized content discovery experience. The result was a robust recommendation system that not only streamlined content navigation but also laid the foundation for improved user satisfaction and engagement.
Data Preprocessing
To build a robust recommendation system, we implemented the following data preprocessing steps:
Cleaning Video Descriptions: removed irrelevant data, such as social media links and promotional content, to focus on meaningful video context.
Enhancing Category Definitions: to ensure that the recommendations were as accurate as possible, we refined WKO’s categories and created detailed descriptions for them, ensuring the AI could understand the context and match videos to user preferences more effectively.
Language Adaptation: all prompts and analyses were conducted in German to ensure alignment with the linguistic and cultural context of WKO’s audience.
Techniques Used
OpenAI API: leveraged the OpenAI API with tailored prompts in German to ensure the AI comprehended the specific content and context of WKO’s videos.
Multiple Queries: to enhance the accuracy of video categorization, we ran multiple queries with varying parameters, using different combinations of video data, such as titles, descriptions, and tags. This approach allowed for a more comprehensive and nuanced classification of the videos.
Results
By categorizing videos and creating a recommendation-ready system, WKO achieved:
Structured Content Organization: videos were systematically classified, simplifying content navigation.
Personalized User Experience: the recommendation system connected users with videos that matched their preferences, enhancing engagement.
Project duration:
1 week
Team
2
2 Data Scientists
Technologies
Python, OpenAI
Tech challenge
Crafting precise prompts for accurate AI interpretation.
Ensuring recommendations matched video content.
Developing a reliable personalized recommendation system.
Solution
We leveraged OpenAI’s AI capabilities to create a personalized video recommendation system, which helped WKO connect users with the most relevant content. This solution streamlined content discovery by offering tailored recommendations that directly aligned with user preferences, boosting engagement and satisfaction.
