...
The initial screen, where the user selects a language, textbook and chapter and can start reading right away, is efficient for new users , because they don't need to go through the registration process. However, it may be less efficient for experienced users, especially if what they have studied differs from a particular textbook and chapter, because they would then have to select in the "known grammar"/"known vocab"/"known kanji" the parts of the language they do or don't know, every time they visit the website again. Using client-side storage to keep track of this information, or implementing an optional login system which stores such information server-side, would help alleviate this issue.
Another issue with efficiency on the initial screen is that a selector is used to locate the textbook, which, while necessary if the user is able to recognize but not recall their textbook name, may be less efficient than typing a name with autocompletion for users who already knows the name of their textbook .
The choice of a selector box for adding grammatical tags to sentences, while easily learnable because all of the grammatical tag types are visible, may be inefficient for users who already know the category the word belongs to. While the grammatical could be sorted alphabetically, a user might know the particular grammatical pattern under a different name from the terminology that is used in this website. Hence, it may make sense to also allow a search box with autocomplete to be used for searching for particular tag types. Optimally, users shouldn't even have to input what types of grammatical structures in a sentence, but this would unfortunately involve complex NLP work for detecting grammatical patterns in sentences.
Visibility
This design has very little state - the type of grammatical structures and vocabulary that the user is expected to understand are listed in the left sidebar, and this dictates the sentences that will be displayed to the user. While information about the sentence's vocabulary, grammar, romanization, and translation are hidden in the accordion menu by default, this is by design, because it is meant to be relied on only if the user has trouble reading the sentence without assistance.
Error Handling
The usage of opening a new tab for displaying additional sentences when the user clicks the "more sentences" button in the Vocab or Grammar accordian menus, instead of modifying the state of the current tab, is chosen so that it is easily undoable - the user can simply close the newly opened tab to return to the original sentences he was looking at prior to clicking the button.
The editing functions - namely, adding and removing grammatical tags from sentences, and changing translations for sentences, are unfortunately not undoable. For this reason, because this model functions much like a wiki - in that any user is able to edit or contribute any sentence - it may be necessary to introduce some form of history into the interface should accidental editing or vandalism become a problem.