A review-processing API must distinguish overall opinion and opinions about specific aspects where supported. Which capability is most relevant?

1. Speech synthesis
2. Language detection alone
3. Document layout extraction alone
4. Sentiment analysis with opinion mining