SBIR Phase I: Qalaxia: skill-aware query engine for K12 Classrooms

这个SBIR第一阶段项目为K12构建了一个新的查询引擎，解决了现有自动问答系统中的四个主要缺陷。这些缺陷阻碍了学生和教师充分利用这类系统的力量。现有的系统1)不能处理原始的学生查询，比如十辆福特嘉年华售价14.6万美元。每辆福特嘉年华的成本是多少？2)没有考虑学生的技能水平，以返回他们理解的范围内的结果；3)没有社区审查和反馈机制，允许系统了解他们自动回答的质量；4)丢弃从学生提问中获得的宝贵见解，这些见解可以用来指导未来的课堂教学。免费使用的查询引擎解决了所有四个缺点，改善了课堂教学，缩小了成绩差距。它使所有能力的学生和来自所有社区的学生能够从人工智能技术获得自动答案，并从用户社区获得定期答案。平台上的用户社区从学生到教师，再到履行企业社会责任的行业专家。该平台符合联邦和州隐私法规，预计将从寻求在K12社区中建立品牌的公司和EdTech公司获得可观的年度经常性收入。这个项目开发了新的技能感知人工智能技术，可以自动将学生的原始查询映射到主题、子主题和意图中。该技术考虑了用户的技能水平，并自动构建从最相关的教育资源中提取的答案，并对教师自己的资源给予最高偏好。如果检测到的主题属于代数，并且查询的目的是“我如何解决这个问题？”，该技术会做更多的工作。它动态地生成额外的脚手架指令来解决问题。这种方法扩展到了代数之外的学科。此外，第一阶段的研究利用机器学习、自然语言处理和主题检测的最新进展来提取和跟踪学生的技能水平，以及学生的优势和劣势，以供教师使用。最后，该项目还将展示该技术在提出的问题、产生的相关回应及其在代数数学结果方面的影响方面的有效性。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This SBIR Phase I project builds a new query engine for K12 that addresses four major shortcomings in existing automated question-answering systems. The shortcomings prevent both students and teachers from fully leveraging the power of such systems. Existing systems 1) cannot handle raw student queries like - Ten Ford Fiestas cost $146,000. How much does each Ford Fiesta cost?, 2) do not take into account students' skill-levels to return results that is within the scope of their understanding, 3) do not have a community vetting and feedback mechanism that allows systems to learn the quality of their automated answers, and 4) discard valuable insights obtainable from students' queries that can be used to inform future classroom instruction. The query engine, which is free to use, addresses all four shortcomings, improves classroom instruction and closes the achievement gap. It empowers students of all abilities and from all communities to obtain automated answers from the AI technology and regular answers from the community of users. The community of users on the platform range from students to teachers to experts from the industry fulfilling their corporate social responsibilities. The platform is compliant with federal and state privacy regulations and is projected to earn significant Annual Recurring Revenue from corporations and EdTech companies seeking to build their brand among the K12 community. This project develops new skill-aware AI technology that automatically maps student's raw queries into topics, sub-topics and intent. The technology considers the user skill-levels and automatically constructs an answer extracted from the most relevant educational resources, with highest preference given to teachers' own resources. The technology does something more if the detected topic falls under Algebra and the intent of the query is 'How do I solve this?' It generates, on-the-fly, additional scaffolding instructions for solving the problem. This approach extends to disciplines beyond Algebra. In addition, Phase I research leverages the latest advancements in Machine Learning, Natural Language Processing and Topic Detection to extract and track skill-levels, as well as strengths and weaknesses of students for teacher's benefit. Finally, the project will also demonstrate the efficacy of the technology in terms of queries made, relevant responses generated and its impact in terms of Algebra math outcomes.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.