This case study highlights the groundbreaking collaboration between the British Board of Film Classification (BBFC) and Dreamix, a cutting-edge technology company. By leveraging extensive BBFC classification expertise and content metadata, Dreamix and the BBFC have developed an innovative machine-learning algorithm capable of:
- generating localised age ratings for films and digital content covering over 100 countries,
- taking into account their local cultural nuances and sensitivities,
- creating locally recognised results with equivalent accuracies of the local classification authorities
The joint effort showcases the transformative potential of technology in enhancing and streamlining content classification processes. In doing so, this paves the way for more adaptable, efficient, and globally localised age rating systems in partnership with both statutory and non-statutory regulators worldwide.
A Century of Safeguarding Entertainment
The BBFC is an independent, non-governmental organisation responsible for generating statutory and non-statutory age ratings, content guidance and classification of theatrical (cinema) content, physical media and digital content in the United Kingdom.
Established in 1912, the BBFC plays a crucial role in safeguarding the interests of audiences by providing comprehensive and accurate guidance on the suitability of content for various age groups. Utilising a classification system that ranges from U (Universal, suitable for all) to 18 (Adults only) the BBFC meticulously evaluates various content issues including language, violence, nudity, and other potentially sensitive content, ensuring that viewers are well-informed and protected from potentially harmful material.
Committed to evolving with societal and technological changes, the BBFC continuously reviews and adapts its published Classification Guidelines every four to five years through an extensive public consultation process, promoting a balanced approach to content regulation and fostering a culture of responsible media consumption.
The Opportunity
BBFC Compliance Officers review each piece of content submitted to them for classification and generate rich compliance metadata, identifying issues, strengths of issues, and preparing calibrated metadata. Using this metadata, they classify each piece of content into a specific age rating category for the UK.
The BBFC, in discussion with its global peers and customers, identified the challenges faced in the classification of content for global direct-to-consumer video-on-demand platforms – especially in providing locally recognised and locally sensitive content classifications.
Using its archive of rich compliance metadata, the BBFC saw an opportunity to use advanced Machine Learning AI to profile every piece of content and from those profiles learn and generate locally sensitive age ratings. The vision of the AI was to permit the BBFC and other regulators worldwide to generate standardised compliance metadata and generate locally recognised content classifications that adhere to local values.
In delivering this Machine Learning AI platform, a global solution can be offered to VoD operators by regulators ensuring that classification processes are both globally scalable and sensitive to the countries in which they operate.
The Dreamix Solution: Automating Age Classification
Initial prototype and idea verification
The BBFC reached out to Dreamix for consultation about the feasibility of using their archive of classification metadata to train a Machine Learning AI to generate reliable age ratings across multiple countries.
After initial data exploration and leading-edge AI research, Dreamix created a set of advanced Machine Learning algorithms that generated age ratings with high accuracy for each country, this validated the concept. As part of the prototype, a demo portal was created where BBFC employees can manually transfer the metadata for a piece of content and generate age ratings for different countries.
The BBFC used this working prototype to conduct an independent verification of the reliability of the generated ratings using new data, previously unseen by the AI, which in turn came up with the same high-accuracy results. This validation was enough for the BBFC to greenlight the investment of turning the initial prototype into a production-ready commercial product.
Patent application
Immediately after the verification of the initial prototype, Dreamix started work on improving the Machine Learning models for territories with historical data available. The resulting methods were determined by the BBFC and external consultants to be original, novel and innovative enough and the achieved results impressive enough, to justify applying for patents. The BBFC and Dreamix then worked alongside patent attorneys to prepare and iterate several rounds of improvements of a patent application and go through the application process with several patent authorities, including patent offices in the USA, EU, UK and several countries in Asia.
External demo, extending the system and production-ready solution:
Along with the patent application, Dreamix also worked on extending the initial prototype in several significant ways. The tasks for this stage included:
- preprocessing and including additional data in the model
- optimising the model for the increased amount of data
- extending the model to predict not only age rating but also offer local content warnings (specific for each country)
- developing an interface into the AI to allow the product to be delivered as “Software as a Service” (Saas) to other regulators
- utilising extended metadata to improve further the accuracy of classifications and consider any local sensitivities for each country
- creating a production-ready infrastructure for running the solution as a service to other regulators
- designing an external web-based demo that the BBFC can use to show the solution to partner regulators and global content providers
The Results
Following its collaboration with Dreamix, the demo has generated industry interest, and the BBFC is now trialling and partnering the platform with other regulators and global content providers.
In a significant milestone, the United States Patent Office has granted official approval to the patent application, and other patent approvals are imminent.
Based on the above results, the BBFC approved additional investment to further develop the solution improving accuracy, country coverage and the generation of additional localised content advice. Moreover, the BBFC and Dreamix are actively engaged in exploring the potential integration of further machine learning methodologies, large language models and generative AI with the promise of exciting advancements and more opportunities for collaboration and innovation.
