---

Wikipedia explains that ABSA provides several results for different features mentioned in a review rather than yielding one scalar value for the entire document / post. Aspect-based sentiment analysis is to the classic sentiment analysis what the colour TV is to black & white TV.

Consider a fragment of text like, “The breakfast was a bit tasteless but the hotel is close to the major attractions”. If the hotel owner wants real actionable intelligence, they need to know that the sentiment towards the breakfast is negative, while the sentiment towards the location is positive. A single scalar value like 0.14 or -0.57 is completely useless here; no matter how one spins it, it provides no value, and when aggregated with similar vaguely multifaceted posts, will provide a random figure which may or may not be close to reality.

Tisane Labs is the only vendor that provides aspect-based sentiment analysis for over 20 languages and is not limited to a handful of domains.