This post has been republished via RSS; it originally appeared at: Microsoft Tech Community - Latest Blogs - .
In today's educational landscape, the fusion of technology and teaching methodologies has birthed groundbreaking tools that revolutionize the way educators craft and administer exercises and assessments. Among these innovative solutions stands the PolyGlot-Edu Generator, a pioneering platform powered by Microsoft Azure's OpenAI services, and developed by Antonio Bucchiarone, Senior researcher, Fondazione Bruno Kessler (FBK), Trento (Italy) supported by Microsoft geared explicitly towards educators seeking to transform learning experiences through AI-driven exercises.
Breaking Grounds in Educational Innovation
The PolyGlot-Edu Generator redefines conventional teaching approaches by offering a dynamic suite of features. At its core, the tool empowers educators to curate diverse exercise formats effortlessly, ranging from fill-in-the-blanks to multiple-choice quizzes, and open-ended questions. What sets this tool apart is its adaptability, educators can seamlessly integrate their own materials while customizing various parameters like language, difficulty levels, and specific target objectives.
The Engine Behind the Curtain
Driving this innovative tool are the APIs that facilitate various exercise types, leveraging Microsoft Azure's OpenAI services. The tool's prowess in evaluating open-ended responses using the Semantic Kernel further enriches the learning experience. Through the "few-shot prompting" technique, student answers are meticulously assessed, assigning scores based on accuracy levels—0.0 for incorrect, 0.5 for partially correct, and a perfect 1.0 for accurate responses. This approach not only evaluates but also educates by providing correct answers and insightful error explanations.
Embracing Collaboration and Growth
To foster continuous improvement, the PolyGlot-Edu Generator welcomes collaboration and feedback within the Hugging Face space. This dedicated environment serves as a hub for users to engage, offer insights, and cast votes on desired features. By encouraging active participation and constructive feedback, the tool evolves to support a broader spectrum of exercises, including evaluation rubrics, matching questions, and discussion prompts.
Shaping the Future of Education
User engagement remains pivotal in refining and advancing this AI-infused educational tool. By inviting users to experience the PolyGlot-Edu Generator demo and actively participate in the survey, the team behind this initiative gathers structured feedback, gauges user satisfaction, and gains insights into real-world applications. This invaluable information fuels ongoing improvements, ensuring that the tool aligns with educators' evolving needs and continues to shape the future of education.
The PolyGlot-Edu Generator isn't merely a tool; it's a catalyst for transformative teaching methodologies, harnessing the potential of AI to elevate learning experiences, foster collaboration, and empower educators worldwide. Join the movement, experience the future of
education today, and be a part of shaping its evolution.
The Azure OpenAI Services powered PolyGlot-Edu Generator provides educators with a tool for creating different types of exercises and corresponding solutions. Educators can provide educational material directly in different formats, including simple text, links to PDF slides, Word documents, or specific webpages. The tool allows the generation of several types of exercises, such as fill in the blanks, multiple-choice quizzes, and open-ended questions.
Educators can personalize exercises by integrating their own materials and adjusting various configurable parameters, including intended target, language, difficulty level, and more, allowing the generator to tailor the exercise to their specific needs.
With a valid request the tool employs Azure OpenAI Services and in particular Semantic Kernel in tandem with a Large Language Model (GPT3.5). The Semantic Kernel fills a predefined prompt with the provided parameters for the LLM. The generated responses are then perfected by custom functions and finally presented in a consistent format for ease of further processing.
The tool is based on a set of APIs
(https://skapi.polyglot-edu.com/swagger/index.html) to support different types of exercises (i.e., fill-gaps, open answer, multiple-choice quizzes, etc..). For some of them (i.e., open answers) the tool relies on the Semantic Kernel to check the accuracy of the student answers. In particular, it employs the "few-shots prompting" technique, evaluating the answer at three possible levels: 0.0 if it is completely wrong, 0.5 if it’s partially correct and 1 in the case it is completely right. If the answer is not evaluated 1 the API provides the correct answer and explains the errors.
The PolyGlot-Edu Generator
Is a cutting-edge tool for educators seeking to enhance teaching methodologies with AI-driven exercises and assessments. Its advanced features, coupled with the seamless integration of the Semantic Kernel and LLM, deliver a comprehensive and user-friendly experience.
Feedback and Testing
We are sharing our tool in the Hugging Face space (https://huggingface.co/spaces/polyglot-edu/generative-ai-for-ed) to collect feedback, foster collaborative improvement, and extend its features to support a wider array of exercises (e,g, evaluation rubrics, matching questions, discussion prompts).
We encourage you to provide feedback after tying the PolyGlot-Edu Generator demo.
Please vote on features and leave constructive comments to help us refine the tool using the following survey (https://forms.gle/sPAWMTuTweqL2tY8A) , which delves into aspects like user satisfaction and the impact on your intended applications. The survey provides structured insights for ongoing improvements.