As I was looking into SM-2 today and other algorithms, I decided to take a look at what SuperMemo was discussing with what it calls “Incremental Reading”. According to the SuperMemo Web Page, there are five basic steps to Incremental Reading.

  1. importing articles to SuperMemo
  2. reading articles and decomposing articles into manageable pieces
  3. converting most important pieces of knowledge to question-answer material
  4. review of the material to ensure a good recall
  5. handling of the unavoidable overflow of information

The more I read the more I got a feel for what incremental reading was, and what it was intended to accomplish. I also realized immediately that it could be improved upon. Let’s take a look.

Based on Classroom Instruction

At it’s core, the purpose of classroom instruction — in fact the very purpose of a teacher — is to select material which is appropriate for the students to learn. I don’t mean appropriate in terms of General Audience, Parental Guidance, etc. I mean in terms of what the students are ready to learn. Incremental reading is designed to take the place of a teacher in this regard. The basic idea is that you import hundreds or thousands of articles and that an algorithm — similar to the SRS algorithm maybe — can be used to help you find articles you “need” to read first, before reading more difficult or boring articles. It’s a great idea, actually.

But the flaws are immediately obvious. One may ask, how does a student know which articles to add to this system? Further, can we trust the student to come up with properly formed sections from that article? Again, while it does a good job in a way, it is simply not usable for classroom instruction, and having some sort of teacher to actually select articles and prepare material would be an incredible improvement over the current system. Yet, of course, it is assumed no such teacher exists, hence the current algorithm.

But of course, with Kongzi, we do in fact aim to provide this kind of instruction. It is interesting in that many students may in fact want to choose their own articles — obviously for self-study purposes. Perhaps that is fine and we can just provide starting material.

Very Similar to the Mulcher App

The entire goal of the Story Mulcher app we have, is very similar to what SuperMemo does with incremental reading. Users submit Stories, which are then put through a grading and review process. These stories are then presented to students depending on an analysis of the story’s target vocabulary compared to the student’s known vocabulary. For example, we aim to allow a user to say “I need to learn these ten words,” and then come up with a story that has their entire known vocabulary plus only some of or all of the extra words on that list. This step-up approach doesn’t really have a name but is a sort of frequency list targeted vocabulary approach.

The main difference besides the user not needing to interact with the program (just read the articles) is that no flashcards are provided. However, the system is intended to be able to create dynamic Cloze tests based on the students known and unknown target vocabulary.

The Future of the Mulcher App

Reading about incremental reading has certanly affected my plans for the mulcher app. I plan to turn it into a per-user story library similar to the incremental reading approach, but also allow users to import “suggested readings” from a sort of “Kongzi Readings Library”. There will be suggested cards to go along with these suggested readings. The user will also be able to enter their own articles and suggest their own facts for flashcards. In a way this is sort of similar to an idea I had about blog-posting things like songs, etc. or otherwise, “lessons” and having a clickable “flashcard pack” one can add to their deck, directly from the blog/wiki/lesson post. So we don’t really need to let the user maintain their own library of lessons; on the other hand, there are some aspects of incremental reading which require it and suggest we should.

For example, what if the user was able to “rate” the paragraph he read as “can’t understand,”, “difficult or confusing,”, “read again”, “interesting”, “i learned something”, “well understood” etc. and the incremental reading algorithm then chose those articles to be read more often — almost as if the entire article itself, was a flashcard?

At any rate there are some good ideas here, but it certainly can be done by anyone without using “Incremental Reading”. Anyone can read an article for themselves and then enter their own facts into a program like Anki or Kongzi. In that sense SM’s incremental reading feature feels more like a user interface convenience. Still, it is noted that this feature has major potential to become a central, important feature of the prorgam on par with flashcards themselves. It just needs to be worked on, hammered out a bit more.

By Serena

Leave a Reply

Your email address will not be published. Required fields are marked *