Gone are the times of Mad Men when marketers and advertisers used to
bring in focus groups to ask what people think and feel about a
particular product or service. Today, they just need to listen and
pay attention to what people are posting and tweeting on social
This is the reason why modern businesses, governments institutions,
news agencies, and advertising agencies are so interested in
Let’s take a look at what sentiment analysis is, how it works, its
challenges, and why you should care about it:
What Exactly is Sentiment Analysis?
Sentiment analysis, or opinion mining, is a Natural Language
Processing (NLP) technique used to determine the inclination of
people’s opinions (Positive, Negative, or Neutral) within online
unstructured text. Businesses use it to measure brand health and
gain information about a consumer’s perception of a product or
The food at the restaurant was delicious. (Positive)
The service at the restaurant was horrible. (Negative)
The quality of utensils at the restaurant was average. (Neutral)
Approaches to Sentiment Analysis
Sentiment analysis incorporates Natural Language Processing (NLP)
and machine learning techniques to assign sentiment scores to
entities, topics and themes in a sentence or phrase.
Let’s dig in a little bit deeper and take a look at the three most
common approaches to sentiment analysis:
- Rule-Based Sentiment Analysis
A rule-based system applies human-created rules to make deductions and
perform automated actions. It is one of the simplest ways to approach
Here’s how sentiment analysis is performed through a rule-based
Each text is broken down into its component parts (sentences,
phrases, parts of speech).
- Next, the sentiment-bearing phrase or component is identified
A sentiment score (usually between +1 and -1) is assigned to each
phrase or component. For example, the word ‘awesome’ might be scored
+0.7, whereas the word ‘awful’ might get a score of -0.7
Finally, the scores are combined and the text is categorized as
either positive, negative or neutral
Sounds pretty simple, right? However, our experience with human
language has taught us that it’s anything but simple. Human beings do
not speak or write the language following simple rules, and therefore,
it’s difficult to categorize sentences as just positive or negative
Consider this sentence, ‘Surprisingly, the match was boring’. The word
‘surprisingly’ might give a positive score and the word ‘boring’ would
be scored negatively, making the score tilt towards neutral although
it’s clearly a negative sentiment being expressed.
What’s more, the human language constantly evolves. Words acquire
different meanings. Trends change. A rule-based system would require
someone to continuously update terms and phrases which can become
quite hectic and time-consuming.
It should be remembered that rule-based systems are not part of AI or
machine learning. The difference is that an AI system can learn and
adapt. Unlike a rule-based system which is dependent on human
instructions, an AI system can determine the next course of action
- Sentiment Analysis with Machine Learning
Machine learning, or artificial intelligence, uses a set of techniques
that rely heavily on statistics and probability to detect complex
The primary role of machine learning in sentiment analysis is to
enhance and automate text analytics functions, such as parts of speech
tagging. For this, a large volume of text with pre-tagged examples has
to be fed into the machine learning model. Once that happens
successfully, the model is then able to identify different words in
their context. For instance, one will be able to learn that the words
‘Sick Burn’ mean very different things in the contexts of gaming and
Although machine learning is great at comprehending larger patterns,
it can miss out on some less common ones.
- Hybrid Sentiment Analysis
Hybrid sentiment analysis makes up for the shortcomings by combining
the best of both machine learning and rules-based approaches.
The hybrid approach uses statistical models and augments them with
human feedback. This significantly enhances the accuracy of sentiment
Challenges in Sentiment Analysis
Sentiment analysis is still not an accurate science. Sometimes, even
humans are not able to identify each other’s emotions accurately.
There are still challenges in sentiment analysis that need to be
addressed. Let’s take a look at a few of them:
Any word, phrase or sentence should be analyzed according to their
context. Without context, or a previous reference point, it is
difficult to interpret someone’s feelings or opinions.
For instance, it’s difficult to get the true meaning of a quote from a
movie one hasn't watched.
Sometimes, the literal meaning of the text isn’t what was intended by
the author. Sarcasm can be very difficult for machines to understand
or detect. Irony and sarcasm are active research areas in the field of
- Comparison of Two Entities
Whether a sentence is positive or negative depends on whose
perspective it is being looked from. When two or more entities are
mentioned, it can be difficult for a machine to identify the primary
While some emojis are universal, others can mean different things in
different contexts. What’s more, different people use emojis
It is very difficult for a sentiment analysis system to accurately
detect emotions from emojis.
- Categorization of Neutral Statements
Defining neutral is also a hurdle in accurate sentiment analysis. What
constitutes a neutral statement? Is it the one in which there’s no
emotion being expressed? Or the one which contains irrelevant
information? Or the one where the system can’t decide if it’s positive
Why Is It Important for Brands
Despite its limitations, sentiment analysis has become strategically
important for modern-day businesses and marketers. With sentiment
analysis tools, such as SocialPie, companies and
institutions can comprehend the sentiment around their brand, prevent
social media disasters and form actionable future strategies.
Let us take a look at a few reasons why brands should get interested
in social media sentiment analysis:
Get A Deeper Understanding of Audience and Connect Meaningfully
Sentiment analysis helps brands and marketers be in tune with the
pulse of their audience. The better you understand them, the better
you will be able to form your marketing campaigns. Understanding your
target customer’s interests, preferences, and passions will help you
stay one step ahead of your competitors.
Sentiment analysis tools also allow you to connect with your audience
in the language or combination of languages they speak. For instance,
SocialPie, a social listening and intelligence
platform, can help you engage meaningfully with your customers who are
posting or tweeting in many languages, including Urdu and Roman Urdu
Improve Customer Service
Sentiment analysis allows you to look at what customers are saying
about your product or service.
Let’s say, you have recently launched a new product and you want to
know what your customers feel about it. A sentiment monitoring tool
can quickly alert your customer support team if there are any issues
they should know about.
Learning about these issues in real-time can help the customer support
team address customers’ concerns immediately and deal with them
We all know how devastating a single negative review gone viral can
be. Social media platforms, review websites, and discussion forums are
replete with people’s opinions about products and services.
With a sentiment analysis tool, you can track mentions of your brand
on the web. This helps brands identify any brewing negative
sentiments, address them smartly, and avoid a social media crisis.
Make Messaging On-Point
When you monitor social sentiment around your brand over time, you
begin to understand the effects of messaging on your social media
These insights help you create better, more relatable social media
posts and tweets for your followers.
Not only do today’s customers tell businesses what they are feeling
about their products, they also do not shy away from giving
recommendations to improve their products and services.
This information can be very insightful and can help businesses make
strategic, data-based decisions.
Whether you are a social media agency or a large multinational
corporation, sentiment analysis should be a key part of your social
Want to learn more about sentiment analysis and how it can help you
rise in the competitive social media landscape? Head over to our
website and or just email us at