Sentiment Analysis - A Help or a Hindrance in Understanding?

01/05/2016 04:23 pm ET Updated Jan 04, 2017

Do you want to know what your employees or customers are thinking? Do you want to better understand their opinions and experiences? It is never been easier because of an emerging technology that is gaining traction, called machine learning, which analyzes trends, identifies issues, and even gauges reactions in just about any area, including health care, politics, law, sociology and general business. Sentiment analysis, one of the most active areas of machine learning, greatly expands the ability of organizations to enhance the appeal of their products and services.

Today's marketplace, whether it is international or local in scope, is intensely competitive. Thousands of products and services vie for consumer attention, with product developers and service providers attempting to tailor their offerings to fill the needs and wants of consumers. Previously, information pertaining to consumer attitudes, opinions and needs were determined through surveys and focus groups. Today, all that has changed and computers are continuing to streamline and expand this process through use of sentiment analysis.

Collecting Information for a Purpose
Using sentiment analysis - or opinion mining, it is possible to scan millions of Twitter comments, reviews, Facebook posts and other forms of social media in order to determine if the comments that people are making about a particular product or service are positive/negative/neutral, for/against, good/bad or like/dislike.

Sentiment analysis can be quite specific. For example, it is possible for restaurant management to find out what the public is saying about a particular food. Hotels can find out what their patrons are saying about their rooms, amenities and services and address any deficiencies that tweets, opinions and online reviews may indicate. Even pharmaceutical companies can use sentiment analysis to know how the public is responding to a specific marketing or advertising campaign, product introduction, disease area or news item.

Machine Learning Is a Value-Added Technology
Machine learning, while enormously useful, is also extremely complex. Human communication frequently involves obstacles to understanding, such as poor spelling, deficient explanation of context and sarcasm. Additionally, people often have difficulty with these factors and miscommunication is a common problem. Since people often have difficulty deciphering these factors, it is exponentially harder for a machine-learning algorithm to accomplish the same task and properly comprehend the nuances of these communications.

Today, all organizations should be constantly evaluating tweets as to whether the sentiments expressed are positive, negative or neutral. For example, a marketing group in a company may be faced with analyzing tweets to understand if customers like or dislike a new and improved version of an established product.

Another obstacle to understanding communications is context, which can frequently change the meaning of words. For example, "cheap" has a positive connotation when discussing price, but is negative when discussing quality.

Recently, Natural Language Processing (NLP) technology, a branch of machine learning, has gained widespread acceptance, given the increases in volumes of information and data that organizations need to sift through to better understand their customer base. There are many versions of NLP software available, which use different approaches, but all these tools make it easier for organizations to incorporate and use these technologies. However, organizations should use caution as the same software can product different results based on who designed the model and which assumptions were used in the model. For example, I recently evaluated a simple company tagline from one of my biotechnology clients using two different NLP models and came up with both positive and negative results!

Nevertheless, in order to ensure value of sentiment analysis, companies need to carefully consider how to select the machine learning software that fits their specific needs and ensure they have the individuals with the skills needed to leverage the software and accurately interpret its results. Many options are available now. In fact, Google, Facebook, Amazon and Yahoo have developed and released open-source versions of the machine learning software that they currently use in their day-to-day operations.

The Way of the Future
Sentiment analysis is here to stay. It provides organizations with a way to evaluate how consumer opinions and attitudes may impact their business, but organizations need to tred cautiously -- and with open eyes -- on which software used, models recommended and people they select for this challenge. For more information, here is a demonstration from MonkeyLearn, a leader in the open-source machine learning space [https://www.youtube.com/watch?v=ZwHTT4pl8Rc].