On the result page of a word question, there is an example sentence at the bottom, which can be extended to a list. Currently for the list there are the following options under Settings -> Sentences -> Order sentences by:
Normal order
Shortest
With audio
Most liked
Currently, the example sentence seems to always be the same sentence for the same word, and the list is always in the same order, depending on the setting, so all the order options are deterministic.
E.g. for 会, the example sentence is always 彼はその会に社長の代理で出席した, which is the first entry for the list in the dictionary entry for 会. However, there are a total of 132 example sentences for this term, so for variety, it would be nice to have some kind of randomization options, e.g.
Random order
Random order (prefer audio)
Or, if fully randomizing the sentence order is too complicated, maybe it could be possible to just pick a single random example sentence, but leave the list as it is.
This is something I really wish I could do, but is actually quite hard to implement.
I cannot go into all the details, but to give an overview - there are a LOT of sentences in renshuu, and even more word/sentence bindings. On top of that, there is a huge range in sentence complexity (length, word use, grammar usage).
Because sentences are used everywhere in renshuu, calling the database for sentence data happens (probably) millions of times a day. For many users, a decent amount of sentences are not overly useful due to them being too difficult, so renshuu tries to put users into "buckets" based on general level, and then those buckets each have subsets of sentences for each word (all sentences are still visible, but 99% of users never go past the first 1-2 pages of sentences).
To randomize them, it would need to either use the current buckets or (a dream of mine) be able to, on the spot, scan the sentences for ones that are an appropriate level for that user. The buckets, though, only contain the 10-20 sentences per word (before switching over the general list of sentences for each word), and so you'd only be randomizing a relatively small set of sentences.
renshuu is not a big service, and I cannot afford/manage more complex systems to keep the sentence queries running fast enough so that a change like this would not drag down load times.
So are you saying the problem would be with adding randomization across the entire list, but would be fast enough if it would select randomly from the 10-20 sentences in the bucket? Or would the random selection from the bucket also be too slow?
If randomization from a bucket is possible, that would probably be enough. My main issue is only on the first few levels of a new term, when you get a lot of questions in a relatively small amount of time, and the repetition is very noticable then. (Especially because I have enabled "Allow a term to be studied multiple times a day" in Vector handling). On higher levels it's not an issue anymore because the questions are spread out wide enough. So a selection from about 10-20 sentences would still be varied enough for that purpose.
Well, the buckets are already slow (the whole system is) - I've lost count of how many times I've tried to optimize the sentence system on the database end, and it just won't get any faster. Might be beyond my skill set.
Randomizing within the bucket may be possible (I can say that it would not cause any slowdowns), but I am not sure how effective it would be.
I cannot promise a fast response, but I'll pin this in my "give it a shot" list, and see what happens in the next few weeks.
