ပြန်လည်လေ့ကျင့်မှုစနစ်၏ သင့်တော်မှု ရှင်းလင်းချက်
ဤဆောင်းပါးရှည်၏ မြန်မာဘာသာဗားရှင်းကို ပြင်ဆင်နေဆဲဖြစ်သောကြောင့် လက်ရှိတွင် OCAT အက်ပ်နှင့်အညီ အင်္ဂလိပ်မူရင်းကို ပြသထားသည်။
To make sure the shadowing feature in Favorites does not quickly overwhelm users with review work, we ran a 30-day simulation of the current review management system.
This simulation uses the same algorithm that the app actually uses, not a separate idealized rule set.
Simulation setup
We simulated a list containing 60 saved sentences.
Each day, after entering shadowing mode, the system automatically generated:
- Review today
- New sentence recommendations
In the simulation, we assumed the user completed exactly the system-assigned Review today and New sentence recommendations every day, without doing extra practice from Other non-today shadowing content.
The daily target for each sentence is 2 reads. In other words, when a sentence shows `0/2`, the system only considers that day's task complete after the user reaches `2/2`.
How does the system control the pace?
The system prioritizes due reviews first, then decides how many new sentences to recommend based on the current review debt.
The current rule is roughly:
- When there is no review debt, recommend up to 5 new sentences
- With 1-3 due reviews, recommend 4 new sentences
- With 4-8 due reviews, recommend 3 new sentences
- With 9 due reviews, recommend 2 new sentences
- With 10 or more due reviews, recommend only 1 new sentence
Review intervals also grow gradually. After a sentence is completed, it returns for review after intervals such as 1 day, 3 days, 7 days, 14 days, and 30 days.
30-day simulation result
During these 30 days, all 60 sentences entered the shadowing system.
By day 21, all 60 sentences had been covered. This means that if a user steadily completes the daily plan, the current algorithm can gradually bring a medium-sized favorites list into practice within one month, instead of pushing everything to the user at once.
Across the 30 days, the simulation produced 504 shadowing records in total, averaging about 8.4 reads per sentence.
The final read-count distribution was:
- 18 sentences were read 10 times
- 36 sentences were read 8 times
- 6 sentences were read 6 times
This distribution is expected: sentences that enter the system earlier go through more review rounds within 30 days, while later sentences have fewer reads but still complete multiple days and rounds of shadowing.
Does the daily workload get out of control?
In the simulation, most days had a fairly stable workload.
The highest daily workload was 12 sentences. Most days required roughly 6 to 11 sentences.
This suggests that the current new-sentence throttling rule can effectively prevent review debt from piling up too quickly. The system does not keep recommending new content blindly; it reduces new recommendations when review pressure rises.
What we learned
The current review management system is suitable as the default arrangement for the first version of shadowing.
It can:
- Prioritize reviews that are already due
- Gradually introduce new sentences when pressure is low
- Show all due reviews instead of hiding review debt
- Bring sentences back at key intervals such as 1 day, 3 days, 7 days, and 14 days
- Gradually bring a medium-sized favorites list into practice within one month
Of course, this is not a fixed plan that is absolutely optimal for every user. People have different numbers of saved sentences, available practice time, and language foundations, so the ideal pace can vary. We will continue adjusting the new-sentence count and interval strategy based on real usage.
The goal of this system is not to make users clear tasks. It is to help users repeat the sentences that are truly worth mastering, with relatively light pressure every day.
