Introducing sentiment analytics

Cuesense just got a major upgrade. From now on all links and text blurbs are analyzed using advanced sentiment extraction techniques and we will be able to sense and extract topics and opinions in English text. If you are interested what your customers or friends think about a topic, person or brand, Cuesense will be able to help.
The new feature grades opinions in 11 emotional dimensions such as anger and surprise. These dimensions are the results of academic research in the area of human psychology:

  • Acceptance
  • Anger
  • Anticipation
  • Disgust
  • Fear
  • Guilt
  • Interest
  • Joy
  • Sadness
  • Shame
  • Surprise

At the core of this new major feature is a natural language processor that can determine the grammatical structure of a sentence, paragraph, and the entire text.
But this is just the beginning. Here is a list of textual cues we use to evaluate and grade a phase:

  • Abbreviations: words like “awol”, “b4”, “thx”
  • Degree adverbs: words like “truly”, “exclusively”
  • Emoticons: “:D” , “:-(“
  • Idioms: phrases like “stand out”, “a cut below”
  • Interjections: words like “hola”, “ta ta”
  • Capitalization: e.g. “GREAT”
  • Negations: words like “never” or “neither”
  • Prefixes: parts of words like “hyper-“ and “under-“
  • Quality words: e.g. “correct” and “elegant”
  • Emotion words: e.g. “cranky” and “merry”

As you can see, Cuesense can understand informal text too, which comes especially handy in evaluating tweets and blog comments.
Early on we rejected the approach of simply grading an entire text article as positive or negative. Too often an article contains areas with both positive and negative commentary. Instead, Cuesense searches for small chunks of text with the strongest emotions or representing the topic of the article the best. These chunks, or as we call them, opinions, will be used by Cuesense in other parts of the service in features we will release very soon.

Tags:

No comments yet.

Leave a Reply