IntelliGrade is a transformative multimodal AI system designed to alleviate teacher workload and democratize access to high-quality writing feedback. Addressing critical challenges in education—excessive grading burdens, generic feedback, and technological inequity—the platform empowers students and teachers through three core innovations:

Multimodal Accessibility: Students submit handwritten or printed essays via photo upload, bypassing typing barriers prevalent in regions like Hong Kong, where handwriting and complex input methods (e.g., Cangjie) dominate. Leveraging OCR optimized for traditional Chinese scripts (92.3% accuracy), IntelliGrade ensures equitable access for non-typing populations, with 95% of pilot users preferring this effortless method.
Teacher-AI Collaboration: The system automates repetitive tasks (grammar checks, error highlighting) using a hybrid AI engine (GPT-4 + fine-tuned BERT), saving teachers 70% grading time. Educators then focus on high-impact feedback, guided by AI-generated insights like class-wide trends (“65% need stronger transitions”) and customizable rubrics (e.g., 40% logic, 30% creativity).
Deep, Adaptive Feedback: Beyond surface edits, IntelliGrade evaluates logical coherence (via semantic role labeling) and creativity (against a 10,000-essay corpus), delivering step-by-step suggestions (“Add a counterargument like in Exemplar 3”). Pilot results show a 40% improvement in post-feedback writing scores and 89% student satisfaction.

Technical Foundation:
Handwriting OCR: Tesseract 5.0 + OpenCV preprocessing (Deployed, 92.3% accuracy).
Logical Analysis: GPT-4 + spaCy SRL (Integrated & Tested).
Dynamic Rubrics: Custom GPT-4 fine-tuning (Deployed in pilot schools).
Class Analytics: React + MongoDB (Fully operational).

Potential impact：
Equity: Bridges the digital divide for handwriting-reliant learners.
Efficiency: Teachers reallocate time to mentoring, reducing burnout.
Engagement: Iterative feedback loops foster student confidence and skill mastery.
IntelliGrade redefines education by harmonizing AI efficiency with pedagogical insight, setting a new standard for inclusive, scalable learning tools.

Problems:
1. Excessive Teacher Workload
Teacher stress correlates with a decline in educational quality due to cognitive overload. In Hong Kong, teachers often spend a significant amount of time grading essays manually, which leads many to report chronic fatigue that adversely affects the quality of feedback provided. Under these time constraints, educators frequently resort to standardized assessments that fail to meet the unique needs of individual students. This approach weakens teacher-student interactions, diminishes learner motivation, and ultimately hinders educational outcomes.
2. Lack of Personalized Feedback
Personalized feedback is essential for improving writing skills; however, research shows that a large portion of teacher feedback tends to be generic and does not specifically address individual weaknesses. As a result, students often find such feedback to be unhelpful and report confusion about how to revise their work. This challenge is further intensified in larger classes, where teachers may not have sufficient time to customize their responses. Consequently, students with vastly different writing challenges may receive the same advice, which stalls progress for both.
3. Limitations of Existing Technologies
In Hong Kong, students often rely on handwriting due to the limited adoption of pinyin, which creates significant usability barriers with existing AI tools. Many students find typing inefficient, particularly when using alternative input methods, which discourages engagement with digital platforms. Furthermore, current OCR systems struggle to accurately interpret handwritten traditional Chinese, often resulting in garbled text and misleading feedback. Tools like Chatgpt require electronic text input, forcing handwriting-dependent students to either type laboriously or forgo AI support altogether.

Hypothesis:
1.Teacher-AI Synergy: Automating repetitive tasks frees teachers to focus on higher-order feedback.
2.Adaptive Learning: LLMs fine-tuned on pedagogical examples can mimic expert tutors, offering context-aware suggestions.
3.Multimodal Accessibility: Supporting diverse input methods, including photo submissions of handwritten work.

none

The IntelliGrade project employs a robust functional architecture organized into three main modules: Multimodal Input, AI Analysis, and Feedback & Dashboard.

Multimodal Input Module: The OCR pipeline includes preprocessing with OpenCV for image alignment and noise reduction on low-quality photos. We utilize a fine-tuned Tesseract 5.0 model, trained on over 1,000 handwritten traditional Chinese essays from Hong Kong students, achieving an accuracy of 92.3%, validated on 300 test samples. The output includes text with positional metadata, allowing feedback to be mapped directly to the original handwriting.

AI Analysis Engine: This module features a layered evaluation system. The surface layer conducts rule-based grammar checks using a modified Grammarly API. The deep layer analyzes logical coherence using GPT-4, equipped with semantic role labeling (SRL) to assess argument flow and detect missing causal links. Creativity scoring is performed by a fine-tuned BERT model trained on a corpus of 1000 graded essays, evaluating novelty.

Feedback & Dashboard Module: The multimodal output overlays error highlights on original essay images using HTML5 Canvas, alongside text summaries with actionable suggestions. The teacher dashboard, built with React and Node.js, provides real-time class error heatmaps, enabling effective batch feedback editing.

Function Point Technical Application Progress
Handwriting OCR Tesseract 5.0 + OpenCV preprocessing Deployed (92.3% accuracy)
Logical Coherence Analysis GPT-4 + spaCy SRL Integrated & Tested
Dynamic Rubric Feedback Custom GPT-4 fine-tuning with rubric weights Deployed in pilot schools
Class Analytics Dashboard React + MongoDB aggregation pipelines Fully operational

IntelliGrade revolutionizes writing education through three pioneering innovations that address critical gaps in accessibility, feedback quality, and teacher-student collaboration.

Firstly, IntelliGrade introduces multimodal, handwriting-first accessibility by eliminating reliance on typing, which is a significant barrier in regions like Hong Kong where students predominantly use handwriting. By enabling photo uploads of handwritten essays, the system leverages OCR technology optimized for traditional Chinese scripts with 92.3% accuracy, surpassing generic tools that struggle with non-Latin characters. This allows students to bypass slow typing, submitting their work instantly via smartphones or tablets. As a result, 95% of pilot participants found this method "effortless and inclusive."

Secondly, IntelliGrade employs generative AI for deep, iterative feedback. Moving beyond basic grammar checks, it utilizes fine-tuned Large Language Models (LLMs) like GPT-4 to analyze essays across three tiers: assessing structural integrity, mapping logical coherence, and scoring creative expression against a corpus of 1000 graded essays. This enables students to receive actionable, step-by-step suggestions, fostering iterative revisions and contributing to a 40% improvement in post-feedback writing scores.

Lastly, IntelliGrade features a teacher-customizable assessment ecosystem. This dynamic rubric system allows teachers to assign weights to evaluation criteria and upload exemplar essays for context-specific feedback. Additionally, a centralized dashboard tracks class-wide trends, allowing teachers to focus on areas needing improvement. This system reduces grading time by 70%, reallocating hours towards personalized instruction while ensuring compliance with privacy laws.
In summary, IntelliGrade bridges technical innovation with pedagogical practicality, democratizing access for non-typing populations and enhancing engagement through tailored guidance.

By transforming how feedback is created, delivered, and refined, IntelliGrade sets a new standard for AI-powered educational tools, empowering teachers and engaging students to foster skill mastery.

To ensure IntelliGrade is scalable and can meet increasing user demand, several strategic measures will be implemented. First, the platform will leverage cloud-based infrastructure, allowing it to dynamically allocate resources based on real-time usage. This flexibility enables seamless scaling during peak times, ensuring consistent performance without latency. Additionally, using microservices architecture will facilitate independent scaling of different components, such as the OCR module and AI feedback systems, allowing more efficient resource management and updates.

To address potential bottlenecks, robust monitoring tools will be established to track system performance and user activity. This proactive approach will help identify and mitigate issues before they affect users. Implementing a phased approach for feature rollouts will also allow us to test scalability in increments, ensuring that updates do not overwhelm the system.
In terms of environmental sustainability, IntelliGrade will prioritize energy-efficient cloud providers with renewable energy commitments for data storage and processing. By optimizing algorithms to reduce computational load, the platform will further minimize its carbon footprint.

To foster long-term user engagement, IntelliGrade will incorporate user feedback loops and adaptive learning pathways, allowing the platform to evolve based on user needs and preferences. Features like personalized dashboards and continuous updates based on educational trends will keep users invested in their learning journeys.

Finally, regular training sessions and community-building initiatives will be established to create a supportive user ecosystem. By cultivating a community of educators and students who share insights, we will foster ongoing engagement and a collaborative approach to improvements. Through these strategies, IntelliGrade will not only scale effectively but also ensure sustainability and adaptability in an ever-changing educational landscape.

IntelliGrade tackles pressing social challenges in education by democratizing access to quality feedback and empowering educators, which aligns with global goals of equity and inclusion.

First, IntelliGrade promotes educational equity by bridging the digital divide for students who rely on handwriting or non-Latin scripts, such as traditional Chinese. Many of these learners face exclusion from AI tools due to inadequate Optical Character Recognition (OCR) capabilities and typing barriers. IntelliGrade overcomes this with its photo-based essay submission system and optimized OCR technology, which achieves 92.3% accuracy, ensuring that non-typing populations, especially in regions like Hong Kong, can access tailored educational support.

Second, the platform significantly enhances teacher efficiency by alleviating excessive grading workloads, which can average between 8 to 10 hours weekly. This overwhelming burden contributes to teacher burnout and diminishes instructional quality. IntelliGrade automates 70% of grading tasks, empowering educators to reallocate their time towards mentoring and innovation in their curricula, thus reducing stress and enhancing their effectiveness.

Third, IntelliGrade actively fosters student engagement through iterative feedback loops, encouraging confidence and mastery of writing skills. By moving beyond generic feedback, the AI-driven analysis of logic, structure, and cultural context offers personalized insights that promote higher-order writing competencies. This dynamic interaction not only bolsters student motivation but also supports continuous learning and improvement.

Aligned with broader objectives, IntelliGrade directly supports UNESCO's Sustainable Development Goal 4 by enhancing inclusive and quality education while promoting digital inclusion tailored for marginalized learners. To stay responsive to community needs, IntelliGrade implements feedback loops with monthly surveys for teachers and students, along with a public feature request portal. Collaborating with primary and secondary schools in Hong Kong helps pilot updates, including support for Cantonese idioms, while an advisory board of educators ensures alignment with evolving pedagogical standards.

Personal Information Collection Statement (PICS):
1. The personal data collected in this form will be used for activity-organizing, record keeping and reporting only. The collected personal data will be purged within 6 years after the event.
2. Please note that it is obligatory to provide the personal data required.
3. Your personal data collected will be kept by the LTTC and will not be transferred to outside parties.
4. You have the right to request access to and correction of information held by us about you. If you wish to access or correct your personal data, please contact our staff at lttc@eduhk.hk.
5. The University’s Privacy Policy Statement can be access at https://www.eduhk.hk/en/privacy-policy.

• I have read and agree to the competition rules and privacy policy.