NTUCB

Website

https://ntu-course-chatbot.ml/

Installation

Create a Python3 virtualenv and run sh install.sh.

We used theano backend, which can be set in ~/.keras/keras.json.

Functionalities

After install, you may access

• python3 manage.py runserver
  • Interactive demo site
    • Rule-based: http://localhost:8000
    • RL: http://localhost:8000/rl
  • Django admin http://localhost:8000/admin
• python3 manage.py shell
  • Making queries in Django interactive shell

Template generation

python3 generate_template.py Generated files will be put in request_template folder.

request_template
├── [   90775360]  classroom.txt
├── [   65659689]  instructor.txt
├── [   87333770]  schedule.txt
└── [   79905920]  title.txt

~1M sentences for each goal.

Current DB schema:

• Course (raw_type: html-format)

  Semester	Serial_no	Designated_for	Curriculum_no	Class_no	Title	Credits	Curriculum_identity_no	Full_half_yr	Required_elective	Instructor	Instructor_url	Sel_method	Schedule_str	Classroom	Classroom_url	Capacity	Course_limits	Remarks	Syllabus_url	Description	Goal	Requirements	Office_hours	Textbooks	Grading	Progress	Course_url
  104-1	73565	網媒所	CSIE5430		機器學習	4	922 U4240	半年	選修	林軒田	http://nol2.aca.ntu.edu.tw/nol/coursesearch/teacher.php?op=s2&td=902083	2	一3,4三3,4	資103	http://map.ntu.edu.tw/ntu.html?layer=&uid=AT3001	176	限學士班三年級以上,本校修課人數上限：176人	初選不開放。本課程將配合開設大型線上開放式課程。	http://nol2.aca.ntu.edu.tw/nol/coursesearch/print_table.php?course_id=922 U4240&class=&dpt_code=9440&ser_no=73565&semester=104-1&lang=CH

• Review (raw_type: html-format)

  Title	Loading	Sweetness	Stars	Content	Sentiment	Probability
  [評價] 105-1 臺灣史一 李文良	0	0	0	（是／否／其他條件）：是 哪一學年度修課： 105-1 ψ 授課教師 (若為多人合授請寫開課教師，以方便收錄) 李文良 λ 開課系所與授課對象 (是否為必修或通識課 / 內容是否與某些背景相關) 歷史系大一必修 δ 課程大概內容 明末/大航海時代~開港前台灣史 Ω 私心推薦指數(以五分計) ★★★★★ 滿天星 η 上課用書(影印講義或是指定教科書) 自編講義，可於ceiba取得 μ 上課方式(投影片、團體討論、老師教學風格) 口述+一點點版書，不過老師聲音頗輕柔，太累的話就QQ惹 σ 評分方式(給分甜嗎？是紮實分？) 好像頗甜，每個星期都有指定閱讀，紮實與否純看用功程度XD ρ 考題型式、作業方式 三篇讀書心得，期末考申論題+名詞解釋 ω 其它(是否注重出席率？如果為外系選修，需先有什麼基礎較好嗎？老師個性？ 加簽習慣？嚴禁遲到等…) 全簽，助教課要到，正課好像是隨老師心情點名 Ψ 總結 如果每個星期讀書的話其實會覺得正課有點無聊，然後期末考都好簡單這樣XD 不過依照討論課的情況，真的每星期讀的人好像不多 然後討論課刷存在感好像++++++很多分～ 　　　　　　　 讀書心得給分頗甜 我大概16個星期(18扣期中期末)認真讀了13個星期吧(圖書館大好OwO) 最後A+ 有讀書的話期末考唯一的困擾就是手痠 --	Positive	0.6

References

• Bordes, Antoine, and Jason Weston. "Learning end-to-end goal-oriented dialog." Proceedings of The 5th International Conference on Learning Representations. 2017. (last updated: 2017/4/17)

• Serban, Iulian Vlad, et al. "Generative Deep Neural Networks for Dialogue: A Short Review." arXiv preprint arXiv:1611.06216 (2016). (last updated: 2017/4/17)

• Shah, Pararth, Dilek Hakkani-Tür, and Larry Heck. "Interactive reinforcement learning for task-oriented dialogue management." NIPS 2016 Deep Learning for Action and Interaction Workshop. 2016. (last updated: 2017/4/17)

• Li, Xiujun, et al. "A User Simulator for Task-Completion Dialogues." arXiv preprint arXiv:1612.05688 (2016).
  (github-repo here) (last updated: 2017/4/23)

• Li, Xuijun, et al. "End-to-end task-completion neural dialogue systems." arXiv preprint arXiv:1703.01008 (2017).
  (github-repo here) (last updated: 2017/4/23)

• Yun-Nung Chen, et al. "End-to-End Memory Networks with Knowledge Carryover for Multi-Turn Spoken Language Understanding (last updated: 2017/5/6)