Columnist: Computer voice-recognition technology currently cannot distinguish between homophones such as "their" and ...

Irina on July 13, 2019

@Shiyi The assumption we are looking for is that the voice recognition program must be able to distinguish between homophones to produce accurate translation, and to distinguish between homophones, it needs to recognize and utilize grammatical and semantic relations among words.
In other words:

recognize grammatical and semantic relations -> distinguish between homophones -> accurate translation.

Answer choice D is incorrect because it flips steps 1 &2 saying that in order to recognize grammatical and semantic relations among words, the technology must distinguish between words that are homophones, i.e. distinguish between homophones -> recognize grammatical and semantic relations.

Let me know if this helps.