!!! Friday

!!Before lunch
* Database model details

!!Last session (after lunch):

* Scan in more schemes and add to SVN
* Where to start?
** look at frontend and backend schemes
** action points, order, time, planning
* Start up django project and initial modules in {{icall/oahpa/}}
* Divide the project subtasks and milestones between Tiina, Ryan and Heli
** how much time everyone has
** time span forward
** deadlines for milestones

* future organization (tools: trello, slack, etc.)
* stakeholders (Russian instructors and representatives)
* pilot language: Russian, test languages: North Saami, other Saami languages

!!Milestones

!First goal: Topic -> Exercise rounds: "Leksa" exercise types. 
* Working demo for some part of Russian, and a demo for some part of North Saami

!General steps
* Linguistic database: Lexicon and morphology models
* linguistic database: some unittests for data relationships 
* Linguistic database install: initial word and feature models, with extensibility
* linguistic database install: FST generation of wordforms from lookup server
* Frontend: topic list view, task list (exercises) view(s)
* backend: topic list view data structure
* backend: exercise question and answer format
* backend: exercise submission format
* backend: exercise question generation for "Leksa" with types (text -> text, text -> multiple choice)
* backend: exercise answer validation for these types

!Working on modules, task assignment: 
* (lene: wednesday, sept 28) first learning package for saami
* (lene, trond) looking at leksa logging
* (tiina) backend linguistics database
* (tiina) backend users database
* (tiina) backend configuration - source data, language rules, metadata
* (heli) backend generation
* (heli) backend answers
* (ryan) frontend views
* (ryan) frontend templates
* (ryan) frontend basic css
* (ryan) frontend connections to backend
* (ryan) lookup server copy / integration
* (robert) first learning package for russian
* (lene, robert) start with error FSTs for Oahpa for Numra and lexicon - north saami, russian

!Notes

Don't be afraid to ask people to contribute

* level model must contain 
** task objects for words, grammar to display
** amount of questions to answer
** types of exercises to present 
** percentages of new, old, and missed words to present
* level that is a test

!!Next Meeting

11. Oct 2016, Minnesota: 08.00, Norway/Sweden: 15.00


!!Initial project setup

* create virtual environment
* initialize django structure
* initialize main apps:
** backend
** frontend
* initialize backend sub-apps:
** linguistics database
** configuration
** generation
** answers
* initialize frontend sub-apps:
** topics
** exercises
** pages

!Technical notes

* Python 2.7
* [virtualenv|https://virtualenv.pypa.io/en/stable/]
* [Django 1.8.8|https://www.djangoproject.com/download/] (see LTS notes)
** [Django 1.8 docs|https://docs.djangoproject.com/en/1.8/]
* MySQL / Mariadb

!Starting up with django

virtualenv env

first time only: django-admin startproject oahpa

then, create apps with: python manage.py startapp







!!Sketching things out


!!Database model details

!Database model sketches

* Answer validation models
* Question/Answer generation models
* user interaction backend models?

!Today's schemes:
* [Backend Models: User interaction|schemes/Backend_User_interaction.pdf]
* [Frontend Exercise Interactions for Student|schemes/Frontend_Exercise_Interactions_Student.pdf]
* [Instructor's interface for adding tasks (Jordan's dream)|schemes/Instructors_interface.pdf]
* [Question/Answer generation|schemes/Question_answer_generation.pdf]

!Database model sketch discussion

Answer validation models

* How do we handle parallel forms. Possibilities:
** Form-Form relationship: table containing relationship type, and form connections
** canonical tag vs. detailed tags: {{+V+Prs+Sg1 ~ +V+Prs+Sg1+Orth/Etc}}: use the canonical tags to find all parallel forms, and then the detailed tags to determine the difference

Question/Answer generation models:

* whether or not sentence generation needs to select relevant forms: are the templates for questions overspecified or underspecified for morphology tasks? 


!User interaction models:

* topics should be enough to get back to the 'old' oahpa of specified training



http://giellatekno.uit.no/ped/crkdoc/meetings/150306.html


!!!Next meeting

Meeting time: \\
Minnesota: 0800 -> ( = 15 -> Norw time