Sentiment Analysis

Sentiment analysis answers the question whether the author is positive or negative about something. Instances of detected sentiment are logged under the sentiment_expressions section; polarity determines if the sentiment is:

• positive
• negative
• mixed

Optional Settings

• explain - if true, includes the explanation for flagging
• snippets - if true, includes the fragment responsible for the sentiment
• document_sentiment - if true, the overall sentiment of the entire text is provided at the root-level sentiment attribute.

What Is Aspect-based Sentiment Analysis (ABSA)?

Wikipedia defines ABSA as an approach that identifies sentiment for specific aspects mentioned in a review, rather than assigning a single sentiment score to the entire document or post.

In essence, aspect-based sentiment analysis does for sentiment analysis what color TV did for black-and-white TV: it adds depth and clarity.

Consider this review:

"The breakfast was a bit tasteless but the hotel is close to the major attractions".

A hotel owner looking for actionable insights needs to know that:

• Sentiment towards food is negative.
• Sentiment towards location is positive.

A single sentiment score like 0.14 or -0.57 would be meaningless here. When aggregated across multiple multi-faceted reviews, these types of scores create a misleading picture that fail to capture real customer sentiment.

Example

Request:

{
  "language":"en",
  "content":"The breakfast was a bit tasteless but the hotel is close to the major attractions",
  "settings": 
  {
    "format":"review", "snippets":true, "document_sentiment":true
  }
}

Response:

{
	"text": "The breakfast was a bit tasteless but the hotel is close to the major attractions",
	"sentiment": 0.12345679012345679,
	"sentiment_expressions": [
		{
			"sentence_index": 0,
			"offset": 0,
			"length": 33,
			"text": "The breakfast was a bit tasteless",
			"polarity": "negative",
			"reasons": [
				"tasteless"
			],
			"targets": [
				"food"
			]
		},
		{
			"sentence_index": 0,
			"offset": 38,
			"length": 43,
			"text": "the hotel is close to the major attractions",
			"polarity": "positive",
			"targets": [
				"location"
			]
		}
	]
}