A Decision Science Lens:
Sentiment Analysis
Or whatever your organization is calling it…
Sentiment Analysis, Text Analysis, Opinion Analytic, Social Data Analytics, Natural Language Processing, Text Data Mining, Unstructured Classification Heuristics, Applied Linguistics, Contextual Analysis, Conversational Synthesis… honestly, this entire article could be filled only with names coined to mean Sentiment Analysis. A name that we only picked because it wins the current binary criteria of newest AND most used.
Sentiment Analysis is a process of collecting human feedback — most often in the form of text, classifying its context, and scoring it. Everything else, including the ridiculous number of names and labels, is academic nuance.
How Can You Say That?
Look, I didn’t say — a waste of time… though it most often is. Nuance can be important, in the same way that context is. But from a Decision Science point of view:
Sentiment Analysis is important only if it allows you to turn unstructured customer feedback into meaningful, actionable, and thereby profitable data for your organization.
All the other components are marginal, for now. Plenty of marginal stuff becomes important later… decades later. I hope academia continues to play in the margins, but businesses need to focus on the big decisions.
The Fall of NLP?
Natural Language Processing was likely the first term to really capture all that Sentiment Analysis now describes. It may not have called out intent and context as strongly as this newest moniker, but all the pieces were there.
Like so many other disciplines in analytics, NLP is now encompassed inside Sentiment Analysis (arguably). Was there a need to layer things like this? I don’t know and I don’t think we should really care. If there was a reason to sunset the reign of NLP, it is best described like this:
Memes, emojis, snapchat, and the like all call into question just how “natural” language really is anymore? This also throws a wrench at Text Mining. Sentiment is just as likely to be conveyed by music, images, and video. While this is clearly just a new form of “natural”, it differs in more ways than it doesn’t with core NLP.
More Of The Same
All of this is just to say that Sentiment Analysis is just like every other process in the analytic playbook. Except where it isn’t:
Collection — this analysis deals almost exclusively with Unstructured Data, mainly text but not for long.
Interpretation — content and intention are key. Humans are ironic, sarcastic, cryptic, and sometimes — nucking futs. This is not just keyword mining.
Classification — from a Decision Science angle, this step doesn’t just organize. It is the step that actually creates the value. Do this wrong and your best outcome is some descriptive analytics. Do it right and you may have a figurative gold mine of customer feedback and intelligence.
Done right -
Sentiment Analysis allows a company to utilize diverse customer feedback to make better decisions and take more profitable actions.
Never forgetting our coming AI overlords…
This is their Waterloo. Will they even know what that means? Will they recognize that I have not stated whether they are Napoleon or Wellington? The point being — all the promise of Artificial Intelligence is actually predicated on whether Sentiment Analysis succeeds. I am fairly certain they will at least interpret that line properly… or am I? ;) … wait was that language, too?
No worries. If the AI can’t understand us, that is just more job security for all of us chaotic humans. Or is it less? For now, recognize the real value and complexity of sentiment analysis done right. Understand what a powerful tool it can be for improving the decision-making ability of any organization. But also recognize the true challenges — you must focus on the bottom line! But as always — thanks for reading!
All typos were fully intended to mess with the sentiment analysis folks. jk ;)
Actionable Sentiment Analysis is done here:
