In my own use of spaced repetition, I've found that it is detrimental to learn vocabulary words in isolation. When you learn a new vocabulary word, you should combine that word with other words so that you have a good understanding of the context that the word might appear in in the real world. One of the challenges that comes with this is that you can spot an unknown word in the real world, but it might exist in a very large sentence that is not appropriate for use in a spaced repetition system. Being able to specify a target word in a text and then deconstruct the text to extract a verb phrase, noun phrase, or other type of phrase that contains the target word, but which is of the appropriate length for use in a spaced repetition system, would be extremely useful.
Example
Target word: Neural network
Source sentence: Some teams used neural networks, which are good at processing sequences of information like sentences.
Ideal output: Some teams used neural networks