We are currently navigating an uncharted media landscape, where data changes at hyper-speed. Consumers buy products from brands they believe in, but they are also quick to state their negative opinions on social channels. Add “fake news” by social media bots into the mix, and brand health can be impacted quickly.

A recent article in MIT Technology Review notes: “Automated accounts are being programmed to spread fake news, according to the first systematic study of the way online misinformation spreads.”

Study researchers were “somewhere between surprised and stunned” to find out that “fake news about all kinds of topics travels farther and faster than news that is accurate” – and by a substantial margin. They found that false news stories are 70 percent more likely to be reposted than true stories.

This underlines how rapidly fake news can undermine a brand’s reputation. To quote Warren Buffet, “It takes 20 years to build a reputation and five minutes to ruin it.”

These days, it might take even less than five minutes.

Getting ahead of negative misinformation before it does damage to your brand is essential. Therefore, keeping a close eye on the sentiment of posts about your brand is crucial. But with data coming in at the speed of lightening, what is the best way to analyze sentiment?

The answer, as with much of media analytics, is “it depends.” Do you prefer automated sentiment, hand-coding to detect sarcasm and other tricky language issues, AI and machine learning, natural language processing (NLP), or a combination of these options?

This is the first of a two-part series to help you answer that question.

Sentiment Analysis Evokes Passion

I have attended dozens and dozens of measurement conferences over the years, but only once has passion about analytics erupted into a shouting match that came this-close to a brawl, which almost resulted in a call to security at the convention center. This happened a couple of years ago, and the topic was sentiment analysis.

On one side were analysts who insisted that manual, human coding is the only way to accurately understand brand sentiment; many of them used sampling analysis to deal with the large quantities of data in social media. They scorned automated sentiment analysis for being unreliable, particularly because of its difficulty distinguishing sarcasm.

On the other side, and equally as determined, were analysts who asserted that human coding is impossible, time-consuming, and expensive because of the massive amount of data available. They claimed that automated sentiment, no matter how flawed, was the only realistic way to analyze data; however, some of them also acknowledged that they would add disclaimers about the accuracy of automated sentiment in their reports to the C-Suite.

There was no way to change minds in this fight a couple of years ago. People angrily dug in their heels and refused to budge. But these days, middle ground can be found with advances in AI, machine learning and NLP.

How Does AI and Machine Learning Help Your Brand?

Artificial intelligence and machine learning help you to process large quantities of data, and make it easier to understand what drives sentiment and customer opinions that affect your brand health. You can summarize tens of thousands of stories to recognize context, which is key to determining tone.

Machine learning helps you to comprehend complex human language such as sarcasm, irony, slang, double-positives, and misspellings. Think about words such as funky, which can be negative or positive, depending on the context; or brilliant, which could be sarcastic. Even words like thin can have two meanings – for example, having a thin smartphone is positive, but having thin walls in your apartment is not.

As mentioned in my previous post, Scoring Your Company’s Brand Health, machine learning provides computers with the ability to learn from previous results, without being explicitly programmed. AI attempts to understand high-level human cognition in terms of natural language processing and more.

How Does Natural Language Processing Help?

Have you ever typed in a question for an automated answer on a website for products and services, such as an airline, a mobile phone carrier, or a retail company? This is one of the best examples of NLP in action.

Specifically, NLP is located at the intersection of computer science, artificial intelligence, and computational linguistics. With NLP functionality, computers can smartly analyze and understand how humans speak, performing tasks such as automated question-answering, sentiment analysis, topic segmentation, translation, relationship extraction, and tagging.

To do this, NLP has to clarify language based on the words, the concepts, and the context. This is not easy, particularly in social media where there are often misspellings, abbreviations, acronyms, little or no punctuation, slang, and ambiguity. Sometimes words from other languages are sprinkled into a conversation too.

NLP isn’t new, but its technology is developing swiftly, with the progression of powerful computers, big data, and human-to-machine communications – the so-called Internet of Things. That being said, the technology remains a work in progress and more advances are still needed, because human language continues to evolve, and it is complicated, ambiguous and rarely clear-cut. How many times have you wished that an article, report or study that you were reading was in plain language?


Perhaps you reside in the camp where human coding for sentiment is a necessary evil. No worries, there are tricks and tips to spot trends without reading every single post.

Maybe you stand firm in the camp that relies on automation for sentiment analysis. These days, a powerful tool that determines sentiment based on machine learning, AI and NLP can give you an overview of your brand health and provide you with critical information.

Or possibly, you want a combination of both automation and human coding.

Regardless, you will need to customize the approach that works best for you to build and protect your brand reputation. Your customization will depend on your brand health attributes, corporate objectives, strategies, tactics, messages, themes, competitors, audiences, and of course, your budget.

My next post will look at different methodologies to consider for sentiment analysis, to assist you with understanding your current brand health and help mitigate risks.

If you would like to learn more about AI, download our white paper here. You can learn more about our new social media bot protection program here.