WARNING This idea is based on geek stuff, math and other programming skills. I understand it is not "simple" to create it, but the idea itself IS simple and we all already know it as we are part of those kind of systems in our everyday web browsing on Facebook, Google or others.
SUMMARYCurrently, on GMC, we have a lot of videos. There are videos of many kind and we can search by genre, style and level in a few clicks. But we can get lost pretty easily and it's hard to find out what we should play and what are the videos that suit our tastes.
The instructors currently answer to this problem on the forum constantly replying to the "What lessons exists for that style or technique?" and "What should I practice?" threads. Fine!
But as a coach and teacher, GMC should be able to propose automatically the videos that suit our needs and genre. And GMC ALREADY know what we like and need. GMC already know what is our level and what genre of music we play because we click and bookmark videos we like and we do REC of lessons we like.
And GMC know what we SHOULD practice because we have a bunch of awesome instructors who rate our playing and tell us what techniques we should work on.
But what is missing is a piece of code that build a bridge between the users and the videos database in a intelligent manner giving the members the opportunity to discover new videos they WILL like.
BANK & SOURCE:Today, we have :
GMC videos BANK sorted by
- genre or music category
- techniques (tags, in the style of...)
- levels
Member usage and preferences statistics (SOURCE)
- clicks on videos
- bookmarked videos
- REC played
- level and ratings given by instructors
The idea is to use the SOURCE data to create usage and preferences statistics sorted by genre, techniques and levels and then request the BANK for the lessons that suit those filters.
I separated the idea in 3 different levels of realization, growing from easy to complex.
LEVEL 1 (easy) :The first level base itself on data GMC already have: clicks and bookmarks.
By aggregating those data and with a bit of math, it is pretty easy to guess the current levels, genres and techniques a member is currently watching and practicing and with those data you can propose the appropriate lessons to reach the next level or perhaps more importantly the lessons that a member will appreciate.
To picture the idea, let's say you have mostly rock, legato and sweeping videos in your bookmarks, that the average level of those lessons is 3.6 then GMC can propose rock, legato and sweeping videos of level 4 and 5. Easy ? Yep, it is...
Then, by giving a weight to the source data we can differentiate the important data and the less important one. For example, the videos you click are less relevant than the videos you bookmark, and the bookmarks are less relevant than the videos you actually play within the REC program.
So by giving a weight to every single data, we can sort out the members' preferences, taking the example above :
1. rock (lot of click, lot of bookmarks, 3 REC played)
2. legato (lot of click, a few bookmarks, 1 REC played)
3. sweeping (a few click, more bookmarks, no REC played)
4. Emir's solos (lot of click, a few bookmarks, no REC played)
Now, the system can propose rock lesson with legato, if possible, played by Emir and perhaps a bit of sweeping technique.
This first step is to teach GMC how to correctly read the statistics and the 2 next levels will simply add more information and data to the algorithm to make it more precise and relevant.
LEVEL 2 (medium) :In the second level, we add a new data source: the REC system.
In the REC system, we have the members playing lessons and their playing rated by instructor. So we know what they play (so basically what they like and practice) and we also know if they mastered those genre and techniques, or not.
Giving those new data we can now propose videos we should like and videos for improvement so as a member I can choose within those different propositions the one that suit my current mood (and GMC know how I feel today !
).
LEVEL 3 (godlike) :This is the last level. The idea here is to integrate the system in the MTP and REC by allowing the instructor to access an interface for every member and tell: "Ok, John Doe, you need to practice your legato and vibrato, theory is not perfect and perhaps a bit of sweeping will make you even better. In that precise order".
We can now propose legato and vibrato lessons at first and then sweeping lessons still taking in consideration the fact we play rock and Emir's lessons.
CONCLUSION:Now we have an automatic system that can propose useful lessons for the user.
The whole idea is to make GMC act more as a human teacher and less as a machine by making the site understand the members and give them good advices and recommendations.
We can add a LOT more information to those statistics and greatly improve the system I describe above. Every data is valuable in some way. For example, we all know the "I like" button on Facebook. This is the best way to know the user preferences. We can calculate the time spent watching a video.
Told you it is geek stuff, but in the end it is useful.