Allow to sort texts by %known (not just by %unlinked)
B
BasDe
Allow to sort texts by %known and not just by %new words (which is actually equal to %LingQed) words. %known could help to improve the graded reader experience.
New (or unlinqed) = 0% known, recognized = 25%, familiar = 50%, learned = 75% and known = 100%.
%known is calculated by: sum of percentages for each word / number of words
Mark Kaufmann
BasDe An interesting idea. %New tells you how many blue words are in the lesson. We think this is the most relevant data point in determining difficulty of new content. The Known% might require too much explanation for most users to benefit from it. It's also not clear whether you have a lot of well learned words or fewer less well learned words.
Could be interesting but all of these calculations and sorts are very resource intensive so we are very careful about adding additional calculations like this. I would say this is unlikely to be done for that reason.
B
BasDe
Mark Kaufmann: Hi Mark,
Thanks for your reply.
I think that percentage new is not necessarily the most relevant data point in determining difficulty of new content. It depens on the type of user. For users that study one text thoroughly before moving to the next, I agree that percentage new works.
However, for users that are more nomadic (guessing there is a fair number of those on LingQ) and want to switch to various new texts when they choose to, the relevancy of the percentage new metric breaks down. If you have ten or twenty texts that have zero unlinked words but plenty of words that you don't know at all, the metric of percentage known would be more relevant.
Also, I think that the calculation and sort is only marginally more complex than for the percentage new metric. Instead of 2 factors for the calculation (#new and #total) you'd have 3 factors (#known, %known and #total). Or would this add a 50% load to the whole thing?
I'd be interested in your reply.
Mark Kaufmann
BasDe: Our approach is definitely not one of nailing down one text before moving on. Most users hopefully are moving on to new lessons regularly. The new words percentage is an indicator to help users gauge the difficulty of a new lesson. That combined with the visual display of the new words and LingQs numbers are good enough indicators.
Adding this new number may or may not help. The fact is these are just indicators, you then have to open a lesson and try it. Adding a very heavy transaction in hopes of providing some incremental benefit is just not something we would look at. Even though it is interesting information.
Someday, if we can figure out how to make these resource intensive transactions happen more efficiently, providing additional information like this would be nice. The more stats and info we can provide the better. But, it isn't realistic right now.
B
BasDe
Mark Kaufmann: If resources are such a big factor here (I wasn't aware that it is), then I completely understand it works that way from a cost/benefit perspective. Thanks for taking the time to consider and reply.
B
BasDe
Mark Kaufmann: I agree that missing this indicator is not the end of the world. Still, the topic seems to be drawing my attention as a puzzle. I have only limited experience in the area of software development, but I could see that this might work through matrix operations.
When you have one matrix with all the lessons and the words that are in it, and another vector per user that has all the words with their knowlegde score, a multiplication between the two could work to find this metric. I get the impression that some programming languages would be more efficient in this kind of thing through vector/matrix operations.
Probably not making a difference (could be many reasons why such a thing doesn't work out in certain context), but wanted to mention this idea anyway.
Mark Kaufmann
BasDe: Thanks. We have looked into a variety of solutions including these. The current setup is as well as we could do. But, things change all the time. We are working on some things in the background that could improve performance in the future.
B
BasDe
Mark Kaufmann: Sounds cool! Thanks for the efforts so far.