Quick-Comments Abstract

I looked at your google scholar and was wondering if you have done anything with one sentence explanations?   I am not sure if I talked about the grant funding me to do NLP of students open ended-responce in 7th grade math for the three major open ended book.   
There are three major free textbooks in middle school math (EngageNY (aka Eureka Math), Illustrative Mathematics, and Utah Math).   Some of these books are seeing large uptake;  EngageNY has been used by half of all math teachers (at least pieces, since it free).  Illustrtive Math that was just released looks like it will quickly surpsace eEngageNY judging from the reception we see at teacher conferences.  All three of these books have about 30% of their questions each night be opened-ended.  

We are paying a small cohort of teachers with grant funding to  read their students open-ended responces and to make comments.  The second group of teachers will be given three comments to choose between.  Some of the choice will be the comments teachers made, and other will be our machine learned sequence to sequence generated comments.  We call this process a Active Machine Leanring Paradigm as each night we will be refitting out machine learned models to try to use all the data we get each day.  Obvously we get to learn which suggestons teachers pick but less obslsy we can learn from ever time a teacher decided NOT to send any of three we suggested.  These negative example are very important as initally we have only postive example.  
-picture of indented interface- 

ASSISTments already has all these books available inside of ASSISTments.  We have also have a google Classroom application that integrates the ASSISTments functionality (student get immediate feedback and teachers get grades pushed up into google classroom so they can look at stuff like common wrong answers students are making) .  Last year 300,000 questions were graded by teachers and 50,000 time a teacher wrote a comment back to a student. 
-add photo of the commenting infrastuate that already exists----

We now want to study how to do this at scale.  On goal is to directly give comments back to student real time.  Maybe other groups would take that approach but we think the teacher is key and valuable and we want to first create a feature like Google Smart reply; Google Smart Reply is the feature that gives gmail users three suggestions on how to quickly response to the students 

_add photo of smart reply melanie


Comments