The Vision: Accessible CEFR Certification
Examinizer began with a clear objective: democratize CEFR language certification. The founder aimed to bypass the traditional, often costly, test centers and make proficiency testing available online to anyone, anywhere. This ambition required building a robust, multilingual platform capable of generating and managing a vast number of language proficiency questions.
Technology Stack: PHP, AI, and PDF Generation
The platform's foundation is built on a modern PHP 8.2 backend, leveraging MySQL 8.0 for data storage. The frontend, comprising over 1000 pages, is a static HTML, CSS, and JavaScript implementation, ensuring fast load times and a reliable user experience. For the complex task of generating official certificates, the system integrates Python with the ReportLab library, handling the server-side PDF creation.
A key innovation in Examinizer's architecture is the integration of the Claude AI API. This AI is instrumental in the adaptive testing process and in classifying questions. This allows for dynamic test generation and ensures that the difficulty and content of questions align with the specific CEFR level being assessed. Payment processing for European users is handled via myPOS, streamlining the transaction process.
The Core Challenge: Scale and Localization
The most significant technical hurdle was generating over 126,000 unique questions. These questions needed to cover all six CEFR levels (A1 through C2) and be available in 13 different languages: English (EN), German (DE), French (FR), Spanish (ES), Italian (IT), Portuguese (PT), Turkish (TR), Arabic (AR), Chinese (ZH), Korean (KO), Bulgarian (BG), Czech (CS), and Polish (PL). This massive undertaking demanded an efficient and scalable question generation strategy.
To tackle this, the team employed the Claude Haiku API. Questions were generated in batches, with the AI classifying and adapting them to specific language and CEFR level requirements. Each generated question was then meticulously stored in the MySQL database, tagged with its corresponding language and CEFR level. This structured approach ensured that questions could be easily retrieved and served to users based on their chosen language and the test they were taking.
Adaptive Testing and Question Classification
The use of Claude AI goes beyond mere question generation. It plays a crucial role in the adaptive nature of the tests. By classifying user responses and understanding patterns in errors, the AI can dynamically adjust the difficulty of subsequent questions. This creates a more personalized and accurate assessment of a user's language proficiency, mirroring the adaptive capabilities found in high-stakes standardized tests but delivered online.
The AI's ability to classify questions also ensures content quality and relevance. It helps maintain a consistent standard across all languages and levels, preventing the introduction of poorly formed or off-topic questions. This classification process is critical for the integrity of the CEFR certification, ensuring that each level accurately reflects the defined proficiency standards.
Server-Side Certificate Generation
Upon successful completion of a test, users receive a digital certificate. The generation of these certificates is a server-side process, executed using Python and the ReportLab toolkit. This ensures that each certificate is unique, tamper-proof, and accurately reflects the test-taker's name, score, and CEFR level achieved.
The pipeline for certificate generation involves retrieving the user's test results from the database, populating a pre-designed certificate template with the relevant information, and then rendering it as a PDF document. This process is automated and integrated directly into the platform's workflow, providing a seamless experience for users upon test completion. The choice of Python and ReportLab provides flexibility and control over the final output, allowing for visually appealing and professional certificates.
Payment Integration
Facilitating online payments is essential for a platform aiming for accessibility and commercial viability. Examinizer has integrated myPOS, a payment gateway popular in the European market. This integration allows users to securely pay for their tests using various payment methods, ensuring compliance with regional financial regulations and providing a smooth checkout experience for a global user base.
Future Implications and Accessibility
Examinizer's approach demonstrates a viable model for making standardized language testing more accessible and affordable. By leveraging modern web technologies, AI for content generation and adaptation, and efficient server-side processes, the platform addresses a significant market need. The success of this project opens the door for similar initiatives in other fields requiring scalable, multilingual, and accessible assessment tools.