Skip to content

Create 틀렸쥬_idea.md#115

Open
lsm-holylee wants to merge 1 commit intomainfrom
hollylee_ai_idea
Open

Create 틀렸쥬_idea.md#115
lsm-holylee wants to merge 1 commit intomainfrom
hollylee_ai_idea

Conversation

@lsm-holylee
Copy link

No description provided.

@mossland
Copy link
Contributor

💬 Initial Review – English Version

1. Idea Clarification

  • The concept of automated wrong-answer analysis + AI-generated similar questions is clear and compelling.
  • However, the service form is not yet specified. Is this a web app, a mobile app, or both? Do you envision separate teacher / student portals?
  • “Exam sheet image” – are you targeting only multiple-choice answer sheets (OMR style) or also descriptive answers? The OCR / scoring complexity differs greatly.
  • For MVP you mention “math or English, certain grade.” Which exact grade level and curriculum (e.g., Korean middle-school Math, US SAT, etc.)?

2. Implementation & Feasibility

  • OCR stack: which engine (Tesseract, Google Vision, MathPix, custom CNN) and how will you handle handwritten vs. printed?
  • Auto-grading rule set: do you assume a key is provided by the user, or do you detect answers from the sheet?
  • Question-generation: OpenAI GPT-4, or an open-source LLM fine-tuned? How will you moderate hallucinations and ensure pedagogically sound questions?
  • Infrastructure: Will image upload & inference run on your own GPU servers or call external APIs? A rough infra budget per 1,000 sheets processed would help.
  • Data privacy: Are you planning on local storage, regional compliance (GDPR, PIPA), opt-in data usage for model tuning?

3. Potential Resource Allocation

  • Rough estimate:
    • 1 full-stack engineer for web/mobile front end
    • 1 AI/OCR engineer for prototype models
    • 0.5–1 designer for UX flows & result dashboards
    • Possible external SME (subject-matter expert) for validating generated content
  • Does this align with your target timeline?

4. Monetization & Market

  • Your notes list students, parents, teachers, institutes – who pays first? (B2C subscription? B2B licensing to cram schools?)
  • Any benchmark pricing of similar solutions (e.g., Quizlet Premium, Mathpresso QANDA +)?

5. UX / UI Gaps

  • Please describe the five key screens you foresee:
    1. Login / role selection
    2. Exam upload & answer-key input
    3. Auto-graded result view (highlight wrong answers)
    4. Generated practice set viewer
    5. Progress analytics dashboard

Without these, designers cannot start wireframing.

6. Open Questions to Move Forward

  1. Service platform & target devices?
  2. Exact subject / grade scope for MVP?
  3. Answer-key input flow? (Teacher provides, or OCR extracts?)
  4. Preferred OCR & LLM stack?
  5. Primary revenue stream in year 1?

Please clarify the above so we can turn this into a sprint-ready spec.


💬 초기 리뷰 – 한국어 버전

1. 아이디어 구체화

  • 오답 분석 + AI 유사 문제 생성이라는 핵심 컨셉은 명확하고 매력적입니다.
  • 다만 서비스 형태가 아직 불분명합니다. 웹인가요, 모바일 앱인가요, 혹은 둘 다인가요? 교사용/학생용 포털을 분리할 계획인가요?
  • “시험지 이미지”가 OMR 객관식만 대상인지, 서술형까지 포함인지에 따라 OCR·채점 난이도가 크게 달라집니다.
  • MVP에서 “수학 또는 영어, 특정 학년”이라고 하셨는데, 정확히 어느 학년·어떤 교육과정(예: 중2 수학, 고1 영어, SAT 등)을 상정하셨나요?

2. 구현 및 실현 가능성

  • OCR 엔진: Tesseract, Google Vision, MathPix, 자체 모델 중 어떤 것을 고려 중인가요? 필기체/인쇄체 구분 방식도 필요합니다.
  • 자동 채점: 답안지를 사용자가 업로드해 키를 제공하는지, 아니면 시스템이 정답을 추출·보유하는지?
  • 문제 생성: GPT-4 API 사용인지, 오픈소스 LLM 파인튜닝인지? 생성된 문항의 난이도·정답 검증을 어떻게 관리하실 예정인가요?
  • 인프라: 이미지 업로드·추론을 자체 GPU 서버에서 할지, 외부 API 호출로 처리할지? 1,000장 처리 시 예상 비용 산출 필요.
  • 개인정보·학습이력 저장 위치와 국내외 규정(GDPR, 개인정보보호법) 준수 계획은?

3. 리소스 투입 가이드

  • 대략적인 투입 인력 예시
    • 풀스택 개발 1명 (웹/앱)
    • AI·OCR 엔지니어 1명
    • 디자이너 0.5~1명 (UX, 대시보드)
    • 과목 전문가(프리랜서) 검수
  • 목표 일정과 맞는지 확인 부탁드립니다.

4. 수익 모델

  • 1차 수익 주체가 누구인가요? 학생 B2C 구독, 학원 B2B 라이선스, 학교 API 제휴 중 우선순위를 정해야 합니다.
  • Quizlet Premium, QANDA 플러스 등 유사 서비스 가격대를 참고하셨는지요?

5. UI/UX 요구

  • 디자이너가 바로 와이어프레임을 그릴 수 있도록 핵심 5개 화면을 명시해 주세요:
    1. 로그인/역할 선택
    2. 시험지 업로드 & 답안 입력
    3. 자동 채점 결과(오답 하이라이트)
    4. 생성된 반복 학습지 뷰어
    5. 학습 진척 분석 대시보드

6. 다음 단계 질문

  1. 서비스 플랫폼(웹/모바일) 확정?
  2. MVP 과목·학년 범위?
  3. 정답 입력/관리 방식?
  4. OCR & LLM 기술 스택 선호?
  5. 1년 차 주요 매출원?

위 항목을 구체화해 주시면 바로 스프린트 단위 개발 계획으로 옮길 수 있습니다!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants