DoubtiQ: A Smart and Personalised Platform for Instant and Tutor-Verified Doubt Solving

Abstract

While online platforms have made education highly accessible, providing timely solutions for student doubts is still an ongoing challenge. Most of the commonly used platforms deal with doubts either by allocating tutors manually or by predefining FAQs. These processes usually take too much time, and students\' attempts to find an answer to the same question that others have already posted lead to a waste of their efforts and time. The proposed DoubtiQ platform aims to assist students in finding answers to their doubts regardless of whether they post doubts in written form or upload images with handwritten and printed information. In the case of image posting, it first converts these pictures to machine-readable text by applying Optical Character Recognition (OCR). Afterward, the processed doubt is matched with previous solutions available on the platform based on semantic similarity, and, if a close enough solution is discovered, the algorithm provides it right away. Otherwise, the platform will assign the tutor with the necessary specialization to resolve the question via an administrative channel. The process of searching for similar doubts along with tutor assistance makes it easier to find solutions while avoiding redundant efforts.

Introduction

This work presents DoubtiQ, an AI-powered intelligent tutoring system designed to improve online education by solving the major problem of delayed and ineffective doubt resolution.

In current online learning platforms, students often face slow responses, keyword-based search limitations, and lack of support for image-based doubts. Manual tutoring systems also suffer from repetitive workload and delays. To address these issues, DoubtiQ proposes a hybrid system combining AI automation with human tutors.

The system accepts doubts in both text and image form. Image-based queries are processed using Optical Character Recognition (OCR) to convert them into text. The processed text is then cleaned through preprocessing techniques and converted into semantic embeddings using NLP-based models. These embeddings are compared with a database of previously solved doubts to retrieve matching answers instantly.

If a match is found, the system immediately provides a solution in text or video form. If not, the query is escalated to a human tutor through an admin system. Tutors then provide detailed responses in text, PDF, or video formats, which are verified and stored for future reuse.

The platform also includes user role management (free and paid tiers) and monetization through subscriptions and tutor-generated video content.

From the literature review, existing systems are limited by poor handling of image-based doubts, lack of semantic understanding, and absence of human integration. DoubtiQ overcomes these limitations by combining OCR, semantic similarity matching, AI, and tutor support.

The methodology includes modules such as doubt acquisition, OCR extraction, preprocessing, semantic matching, instant retrieval, and tutor escalation. The workflow ensures continuous improvement as new solved doubts are added to the knowledge base.

Performance evaluation using 4,000 student queries shows strong results:

• OCR accuracy: 88%–91%
• Efficient handling of both text and image-based doubts
• Faster resolution compared to keyword-based systems

Conclusion

This study presents DoubtiQ – a smart hybrid system for solving academic doubts that utilizes several technologies for providing more effective academic support services to the students in the digital world. The platform solves one of the major problems associated with online learning by allowing learners to raise their questions in both written and image forms. Thanks to the implementation of OCR technology, the system is able to convert any image-based queries into processable text format. Students are free to use scanned notes and uploaded photos containing questions. After the text is generated, the system compares them with the already existing answers to previous doubts using semantic embeddings and similarity calculation techniques. This enables faster responses and helps tutors save their time as they do not need to type out the answers that were already given several times. The results obtained in the real-world test of the system\'s functionality confirmed its high efficiency. With high accuracy of OCR algorithms and semantic similarity calculation as well as instant resolution of more than two-thirds of the raised doubts, DoubtiQ appears to be an effective tool. From a sustainability perspective, DoubtiQ comes with functionalities such as management of user roles that ensures premium users get priority customer support services as well as an ad-free environment. In addition, tutors have the opportunity to make contributions through video content and earning from views made on such contents. • Looking ahead, the platform has strong potential for growth. Future development could include: • Multilingual OCR, so students can ask questions in regional or native languages. • Generative AI tools to assist tutors in drafting high-quality explanations faster. • Difficulty-level tagging, to help match doubts with students’ current skill levels. • Personalized analytics dashboards, allowing students to track their progress and identify weak areas. • Gamified learning elements, such as badges or rewards for consistent engagement. In summary, DoubtiQ is much more than an innovative technological solution; it is also a very well thought-out method for handling doubt resolution in e-learning based on the needs of the student. This is achieved by combining automation techniques with professional assistance, thus contributing to better education delivery online.

References

[1] Kochmar, P. Alfter, S. Baur, and T. Zesch, “Automated Personalized Feedback Improves Learning Gains,” Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics (EACL), Valencia, Spain, pp. 598–609, 2017. [2] S. Sani and B. Aris, “Computational Intelligence Approaches for Student and Tutor Modeling in Intelligent Tutoring Systems,” International Journal of Artificial Intelligence in Education, vol. 28, no. 3, pp. 412–435, 2018. [3] J. Horwitz, R. Banerjee, and S. Kulkarni, “Creating an Adaptive Training System: Integration of Student Modeling and Intelligent Feedback,” Journal of Educational Technology & Society, vol. 23, no. 2, pp. 45–57, 2020.. [4] S. Kulshreshtha, A. Sharma, and M. Nair, “Few-Shot Question Generation for Personalized Feedback in Intelligent Tutoring Systems,” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, no. 14, pp. 15732–15740, 2021. [5] R. Joshi, K. Mehta, and P. Verma, “Looking for a Match: Self-Supervised Clustering for Automatic Doubt Matching in E-Learning Platforms,” IEEE Access, vol. 9, pp. 112345–112357, 2021. [6] J. Mathew, R. Thomas, and A. Joseph, “NLP-Based Personal Learning Assistant for School Education,” International Journal of Advanced Computer Science and Applications (IJACSA), vol. 12, no. 6, pp. 210– 217,[18:00, 29/01/2026] [7] S. Mallik and A. Gangopadhyay, “Proactive and Reactive Engagement of Artificial Intelligence in Education: A Survey,” ACM Computing Surveys, vol. 54, no. 7, pp. 1-36, 2022. [8] P. Gupta, A. Sharma, and M. Nair, “AI-Driven Question Routing and Tutor Allocation for Online Learning Platforms,” International Journal of Artificial Intelligence in Education, vol. 32, no. 4, pp. 789-812, 2022.

Copyright

Copyright © 2026 Vaibhav Deshmukh, Siddhesh Chattar, Bhsuhan Dighe, Tejas Sabale, Poonam Bhawke. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.