SUSI AI

Case study: How Acty and PÖFF created a film recommender

SUSI AI film recommender using artificial intelligence

We are thrilled to have collaborated with PÖFF (Tallinn Black Nights Film Festival) to create something truly innovative – the SUSI AI film recommendation tool. PÖFF is the first film festival in the world to use artificial intelligence to help visitors find films and events that suit them best. Our role was to assist PÖFF in creating this technical solution by integrating a custom AI system into their website. This system connects the festival’s database with an OpenAI model, enabling it to provide real-time, personalized film recommendations.

Project Goal

PÖFF is one of the largest film events in Northern Europe, showcasing hundreds of films and numerous special events each year. Given the diverse and extensive nature of PÖFF’s program, the SUSI AI film advisor was developed to provide visitors with an interactive, user-friendly tool to quickly and conveniently find personalized film recommendations. This not only saves visitors valuable time but also helps them plan their festival experience. Additionally, SUSI AI can answer questions about films and events in multiple languages, ensuring smooth and accessible communication during the festival.

How SUSI AI Works?

The technical solution is centered on a chatbot built using OpenAI technology, integrated with PÖFF’s internal database. SUSI AI’s process works as follows:

      • Database Integration: All films and events in the PÖFF program are stored in a STRAP database. The system collects fresh information daily and creates new vectors (coordinates) to represent specific film data.
      • AI Solution: When a visitor asks a question, a vector is generated and compared with vectors in the database. This data is used by the AI to respond to the original question and suggest films that meet the visitor’s preferences.
      • Chat Front-End: The chatbot is available on the PÖFF website through an open-source chat window customized by us to meet the client’s needs, providing a seamless and personalized communication experience for PÖFF visitors.

What We Accomplished

We developed a middleware system that connects the OpenAI API with the STRAP database, creating a seamless integration of data and AI. Initially, all information on films and events is sent to OpenAI, which processes this data and returns embeddings, or vectors, to us. These embeddings allow the AI to generate precise and relevant answers based on user questions. To ensure accuracy and relevance, the system continuously updates with the latest festival data, so the information provided to users is always current.

The background setup also allows us to specify how much of the conversation history AI should take into account, and to fine-tune the accuracy of responses to best match user queries. Additionally, we can adjust the style of AI-generated responses to align with PÖFF’s goals, enhancing the user experience and ensuring the recommendations feel tailored to the festival’s unique context.

How SUSI AI Creates Responses to User Questions

The system relies on two main components:

    • PÖFF’s Dataset: This includes festival films, screenings, and other related information.
    • AI API: A language model used to analyze data and generate responses to questions.

In addition, a middleware layer based on PHP is created to perform two main functions.

When a user submits a query, the following process occurs:

  • Query Processing:
      • The system intercepts and analyzes the user’s query.
  • Handling Repeated Queries:
      • If the user has made several related queries, these are combined into a single query sent to the AI.
      • For example, if a user first asks for recommendations on 5 romantic films and then wants to know the cast of the third film, these questions are merged to provide a more comprehensive answer.
  • Vector-Based Search:
      • The AI returns a vector based on the query.
      • This vector is used to locate the most accurate and relevant data (films, screenings, programs, etc.) in our database.
    • Final Response:

      A query is prepared for the AI with:

      • The original user question,
      • Relevant data retrieved from our database,
      • A prompt: specially designed instructions that guide the AI to respond accurately and in the desired format.
    • The AI then analyzes the query content and generates the final answer, which is sent back to the user.

Extra funding

The idea was fueled by an AI workshop held at Tehnopol Science and Business Park, where nine companies were given the opportunity to develop their AI projects. Our collaboration with PÖFF received €20,000 of the total €269,100 funding, which helped bring the solution to life.

Client Feedback

It is extremely important for PÖFF to have a partner who can meet deadlines. Acty did this wonderfully, remaining professional and fast. We could always trust that any issues would be resolved. We look forward to seeing how this solution supports and enhances the PÖFF visitor experience.
Mikk Granström

CEO, PÖFF

Summary

The SUSI AI film recommender represents a step forward in the film festival world, offering visitors an interactive and personalized experience. We are proud to have made this innovation a reality with PÖFF and hope visitors enjoy this new way of enhancing their festival experience.

Join Newsletter!

Keep up to date with e-commerce through the newsletter
best practices, advice and success stories.
how to grow your e-business.