Using BERT model for parsing and doing bigram or multi-gram

Question

The task I want to do is finding the sentiment of some phrases in a group of sentences (not of all the sentences). For example, I have these sentences:

Ended up deciding that a pebble was best. Less charging, more simplistic ui, buttons seem easier for an older person.

And I want to detect these phrases:

1. less charging
2. more simplistic UI
3. easier buttons

I need something that first detects the phrases, and then identifies their sentiments - they are all positive in terms of smart watches for elderly people.

Does the BERT model solve this problem? Or I have to find another tool/technique?