Sentiment Analysis: Comparing is Fair (2)

In out latest post we spoke about the complexity of analyzing comparative sentences for Sentiment Analysis. Now we're going to show some examples of how Bitext API can handle these types of structures.

Comparison of superiority: this type of sentences establishes a comparison level where the subject, focus of the comparison, has a greater degree than the second entity/concept in the comparative structure: “The iPad is nicer than the iPhone”. In this sentence, Bitext's Deep Linguistic Analysis detects two differente sentiment values, according to the position of the compared terms in the structure:

  • "nicer" (3.00, rather positive score) applied to "iPad", the subject,
  • and also "nicer", but on the other side of the comparison (-3.00, negative score), applied to iPhone.

Comparison of inferiority: this type of sentences establishes a comparison level where the subject, focus of the comparison, has a lesser degree than the second entity/concept in the comparative structure: “The iPad is less nice than the iPhone”. This is the opposite of the previous one. Bitext's Deep Linguistic Analysis also detects two sentiment values:

  • "less nice" (-2.00, negative score) applied to "iPad", the subject, and
  • "less nice" (2.00, this time positive), applied to iPhone.

Comparison of equality: this type of sentences establishes a comparison level which is equal or the same between two entities or concepts: “The iPad is as nice as the iPhone”. In this sentence, Bitext's Deep Linguistic Analysis detects two sentiment values that are the same for both terms:

  • "nice" (2.00, positive score) applied to "iPad"
  • and "nice" (2.00, also positive score) applied to "iPhone".

Thanks to the parse tree below, we detect the comparison structure “more… than” (composed by the adverb "more" and the comparative conjunction "than"), the comparative form “nicer” of the adjective "nice", and the two terms of the comparison: one of them, “iPad", as the subject of the copulative sentence, and the other one, "iPhone", as the object of the comparison.

As we have seen, Bitext's Deep Linguistic Analysis is capable of semantically parsing comparative sentences and assign meaningful values to entities and concepts which are the focus of the comparison, in this kind of structures and in more complex ones. We'll see more examples in future posts.

 

Subscribe Here!