How does a lexicon$-$based approach differ from a machine$-$learning approach in sentiment analysis?

Lexicon$-$based approach relies on predefined dictionaries, while machine learning uses labeled data for training

Machine learning focuses on individual word sentiments$,$ while lexicon$-$based approaches analyze entire sentences

Lexicon$-$based models use neural networks, while machine learning employs rule$-$based algorithms

Machine learning relies solely on expert$-$crafted lexicons for determining sentiments$,$ while lexicon based approach uses labeled data for training