This is a series of articles on fundamental technology underlying social intelligence.
Ever since discussion boards and websites started to solicit the opinion of the public, there have been both scientific and enterprise efforts to mine these online texts and derive meaning from them.
With the global internet generating a continuous stream of content and feedback, sentiment analysis has evolved. The earliest version of Natural Language Processing (NLP) was a simple dictionary look-ups of words that contains a positive or negative numerical value, (for example, ‘Good’ = 1, ‘Bad’ = -1.) The NLP Version 1 is still being used in many technology stacks simply because it’s cheaper and relatively fast. The simple model is converting relevant words into a numerical value, and then averaging the value and representing the value as good, bad, or neutral.
With greater computation and more academic investigation, NLP Version 2 comprises analyzing parts of a sentence by looking at the conceptual linkage between words across multiple dimensions. This is a more precise method in determining the meaning, especially useful in a large body of text, or comparing multiple long-form text for patterns and relationship. This advance technology is often used in legal documents where the vocabularies are standardized and the text comprises other text in reference.
While the advance in NLP Version 2 is a great technical advance, it has very limited application in commerce. For sentiment analysis, the true advance comes from emoticons (Emotional Icons) which we base our Behavioral Index on. With Facebook and other social media platforms, comes the powerful use of Likes to show a preference, whether it be consumers or voters. Out of the Likes button comes Facebook’s six emotional indicators, which we term the behavior index. And this represents an advance in sentiment analysis. The user chooses six motions to express their attitude towards the content or a remark. This feedback is nonlinguistic, independent of language, dictionary bias, or culture. It represents a more immediate understanding of the user’s attitude.
The next version of Sentiment Analysis will no doubt include bio-feedback where smartwatches give a biological evaluation of the content we consumer. Until that day comes, we’ll stick with the Behavioral Index.
To schedule a demonstration of our social data search platform, please use https://go.oncehub.com/RogerDo