en.wikipedia.org/wiki/Chinese_input_method

The article states user-specific training is required before these input methods can be used. That is not true for modern mainstream implementations:

• OCR: Tesseract's official documentation shows immediate OCR use after installation: its basic example is simply tesseract myscan.png out. The docs describe training only as an expert task for adding or customizing language models, not something a new end user must do before OCR works.
• Handwriting recognition: Apple's Scribble support says you can write with Apple Pencil in any text field and it automatically converts your handwriting into typed text. Apple's documentation also notes Scribble is on by default when a supported Apple Pencil is paired with an iPad. Microsoft likewise documents handwriting-to-text features that convert handwriting automatically as you write.
• Speech recognition: Microsoft's current voice-access documentation says Windows 11 voice access uses modern, on-device speech recognition to let users control the PC and author text with their voice. This is presented as a built-in feature, not one that requires a user to enter a mandatory learning mode first. Older speech products sometimes offered optional retraining, but that is different from saying speech recognition generally must be trained before use.

So the sentence is inaccurate because it treats per-user training as a general prerequisite for handwriting/OCR/speech input, whereas current systems are typically pre-trained and usable immediately, with optional personalization in some cases.