Hi,

I know that user guide can be a bit vague on this, but based on discussion I had with NV Access people back in 2012 when these settings first made their appearance and experiences from Chinese, Japanese, and Korean speakers (I speak Korean; now you know half of the story):

These settings control how NVDA reacts to input method editor (IME) interfaces. Certain languages ship with IME’s, and because the well-known languages where IME is frequently encountered (Chinese, Japanese, Korean, collectively called CJK) provide different input experiences (candidates window for Chinese and Japanese, standard QWERTY-style input for Hangul/Korean), Windows will present different input experiences, and thus NVDA must respond differently. Visually, these languages are pictorial characters, and when you type in these languages, a square “composition” window will be used to complete each character/shape. How NVDA can announce these characters is beyond the scope of this thread.

The input composition settings panel consist of:

• Automatically report all available candidates: used when entering Chinese characters where you need to choose from multiple candidates for a given pronunciation or a word. These candidates are presented as a grid as soon as you either type the first few syllables or press Chinese character key (right Control key for Korean keyboards). For example, when entering Chinese characters through Korean IME, the character “su” can have different Chinese character representations such as “water,” “number,” “manual,” and so on. NVDA will announce these candidates when this setting is on.
• Announce selected candidate: used when entering Chinese characters, especially when navigating between candidates. When candidates appear, you can use arrow keys to navigate it.
• Always include short character description when announcing candidates: useful when entering Chinese characters through Korean IME. As I noted above, because a single character in Korean can have multiple Chinese character representations, it is helpful to hear exactly what these characters stand for in Chinese through a short description. To support this, Korean translation of NVDA includes short descriptions for thousands of Chinese characters used in China and South Korea.
• Report changes to the reading string: useful for Chinese input methods where different character combinations can yield different texts.
• Report changes to the composition string: when using IME’s with direct QWERTY-style input such as Korean/Hangul, a square window will be visible as you type. The shapes inside the composition window changes as you type, and NVDA can be set to announce changes to this.

Cheers,

Joseph