Welcome to the Digital Marketing Inner Circle

However, given the volumes of consumer generated data, many of the larger marketing/PR firms have moved towards automated platforms that use Natural Language Processing (NLP) to discover, measure and report on various topics such as a brand, product and even an individual.  Commonly called, online reputation management (ORM) marketers rely on these systems’ to employ algorithms in determining a documents sentiment and tone. In this article, I will attempt to demystify the process, provide context, and offer some concrete examples of how businesses can utilize it.

What is Automated Sentiment Analysis?

Most automated platforms identify sentiment differently but here at SinoTech Group, we use a machine learning algorithm to train our SinoBuzz system, which basically means that we have a training database of sentences and a large taxonomy of keywords indicating negative and positive tone. Automated sentiment analysis will never be as accurate as human analysis, because it doesn’t account for the subtleties of sarcasm or irony. However, according to our experience humans only agree 82% of the time on the sentiment of a given document.

Therefore, the point of automated sentiment analysis isn’t to get 100 percent accuracy but to identify the polarity outliers- a method of determining the sentiment of those documents to the right and left on the intensity curve. The sentiment analysis is typically distilled into a whole number then categorized either into categories: positive and negative; or into an n-point scale, e.g., very good, good, neutral, bad, very bad. In addition, sentiment accuracy depends greatly on what is being analyzed. News articles are harder to analyze than a movie or product reviews and shorter documents are harder than longer ones.

This technology is not only great for monitoring online buzz but can also be used for measuring crisis management, or even gauging the efficacy of an online campaign.

Now words of caution.....

Remember in campaign analytics, we caution our clients not too focus on one metric alone (like CTR or impressions). It is the same with social media analytics, sentiment is just one data point and you should incorporate other data points such as the social media channel, the author or location.

Further, as automated sentiment is never going to be 100% accurate, you, the human, need to have the ability to override the sentiment control in the platform. (Note: when selecting an ORM platform, ensure that it supports the manual override of sentiment, the ability to remove spam and irrelevant results). Using an ORM platform is a question of size. Meaning, if you are monitoring a small number of articles (<25 or so), then the most effective way to gauge sentiment is to simply read them. But if your client that wants you to monitor the “Internet” for articles about their products and a simple Google search shows millions of results?

I would suggest this is time to look into an ORM platform that can automate detection and sentiment analysis. From my perspective, ORM platforms are most useful in monitoring and measuring brand sentiment and competitive landscapes. For example, if I am a hotel manager I am very interested in knowing what people are saying about my property in the social networks and/or BBS’s.

Also, I am interested in what they are saying about the competitor across the road. That said, many of the newer ORM platforms provide in depth analysis of authors demographics, their locations and languages used. Therefore, the use of a ORM platform within a modern marketing/PR firm has become compulsory and the days of manually doing counting posts and cutting and pasting example comments is over. Our clients are demanding more information, greater insights and near real-time data monitoring.