You’re likely familiar with the saying, “Texting is a brilliant way to miscommunicate how you feel and misinterpret what other people mean.” You’ve probably even experienced it directly!
Substitute “texting” with “email” or “online reviews” and you’ve struck the nerve of businesses worldwide. Gaining a proper understanding of what clients and consumers have to say about your product or service or, more importantly, how they feel about your brand, is a universal struggle for businesses everywhere.
What if I told you it doesn’t have to be this way? Or better yet, what if I told you there is a way to better understand the feeling embedded in written communication?
Well, there is! It’s called Natural Language Understanding (NLU). NLU comprehends language. To truly understand, we must know the definitions of words and sentence structure, along with syntax, sentiment and intent – refer back to our initial statement on texting. NLU extends a better-known language capability that analyzes and processes language called Natural Language Processing (NLP). By extending the capabilities of NLP, NLU provides context to understand what is meant in any text.
Sentiment Analysis (SA) takes NLU one step further. Sentiment Analysis identifies whether a message is positive, negative or neutral. Together, NLU and SA generate data that tell the story that businesses and enterprises are dying to understand: what customers think and feel about your brand, product or service. To put it another way, “Natural Language Understanding and Sentiment Analysis are brilliant ways to interpret what other people are feeling via their language.”
Why Sentiment Analysis Matters
For the enterprise, Sentiment Analysis is the Holy Grail. Sentiment Analysis allows you to get inside your customers’ heads, tells you how they feel, and ultimately, provides actionable data that helps you serve them better.
Sentiment analysis is the bellwether of your brand.
Sentiment Analysis Approaches
Whether we realize it or not, we’ve all been contributing to Sentiment Analysis data since the early 2000s. To borrow a phrase from “The X-Files” The data is out there.
Consumers offer sentiment information when they:
- Answer an email asking how you enjoyed your Door Dash
- Provide a rating for an Uber driver
- Write a Google review for a business they love (or not)
- Comment on Facebook or Instagram posts promoting a product
Sentiment Analysis determines the tone or opinion in what is being said about the topic, product, service or company of interest. The first step involves classifying the data itself. The underlying approach determines exactly how this works.
So how do you derive sentiment out of language data?
Machine Learning and Deep Learning
Machine learning applies algorithms that train systems on massive amounts of data in order to take some action based on what’s been taught and learned. Here, the system learns to identify information based on patterns, keywords and sequences rather than any understanding of what it means.
Deep learning is a subset of machine learning that adds layers of knowledge in what’s called an artificial neural network that handles more complex challenges.
For example, say you’re a property management firm and want to create a repair ticket system for tenants based on a narrative intake form on your website. Machine learning-based systems would sort words used in service requests for “plumbing,” “electrical” or “carpentry” in order to eventually route them to the appropriate repair professional. Just as a machine learning system is trained to identify and sort the requests, it would also have to be trained to identify sentiment-related terms such as “urgent,” “serious” and “emergency” versus “small” or “minor” and a positive, negative or neutral rating would be assigned.
Machine learning and deep learning are what’s known as “black box” approaches. Because they train themselves over time based only on the data used to train them, there is no transparency into how or what they learn.
Another approach to sentiment analysis involves what’s known as symbolic learning.
The most significant differences between symbolic learning vs. machine learning and deep learning are knowledge and transparency. Whereas machine learning and deep learning involve computational methods that live behind the scenes to train models on data, symbolic learning embodies a more visible, knowledge-based approach. That’s because symbolic learning uses techniques that are similar to how we learn language.
Using a human-like representation of logic and embedded knowledge, a symbolic approach “understands” words or phrases because it understands their meaning, rather than because of how they are trained based on pattern or sequence matching. For example, where a machine learning approach uses documents pre-labeled as positive, negative or neutral to train the model, a symbolic approach would label words as either positive or negative in order to measure sentiment and further, it would be able to understand the context in which those terms appear.
In fact, context is one of the main challenges associated with sentiment analysis. The same words can have a different meaning depending on the context in which they are used. For example, consider a customer survey that asks both “What do you like about the product?” and “What do you dislike about the product?” The response “Everything about it!” would express completely different sentiment depending on which question the client answers.
A prime example of symbolic learning is chatbot design, which, when designed with a symbolic approach, starts with a knowledge base of common questions and subsequent answers. As more users engage with the chatbot and newer, different questions arise, the knowledge base is fine-tuned and supplemented. As a result, common questions are answered via the chatbot’s knowledge base, while more complex or detailed questions get fielded to either a live chat or a dedicated customer service line.
Sentiment Analysis in Action for Better Internet Banking
Expert.ai’s Natural Language Understanding capabilities incorporate sentiment analysis to solve challenges in a variety of industries; one example is in the financial realm.
Accurately understanding customer sentiments is crucial if banks and financial institutions want to remain competitive. However, the challenge rests on sorting through the sheer volume of customer data and determining the message intent.
Such was the case for Intesa Sanpaolo. When the banking group wanted a new tool that brought customers closer to the bank, they turned to expert.ai to create a better user experience.
Expert.ai employed Sentiment Analysis to understand customer requests and direct users more quickly to the services they need. For example, thanks to expert.ai, customers don’t have to worry about selecting the “right” search expressions, they can search using everyday language.
Because expert.ai understands the intent of requests, a user whose search reads “I want to send €100 to Mark Smith,” is directed to the bank transfer service, not re-routed back to customer service. Only six months after its launch, Intesa Sanpolo’s cognitive banking service reported a faster adoption rate, with 30% of customers using the service regularly.
Other applications of sentiment analysis include using AI software to read open-ended text such as customer surveys, email or posts and comments on social media. SA software can process large volumes of data and identify the intent, tone and sentiment expressed.
Take social media comments for instance. Social media users are able to comment on Twitter, Facebook and Instagram at a rate that renders manual analysis cost-prohibitive. Analysis of these comments can help the bank understand how to improve their customer acquisition and customer experiences.
Read more practical examples of how Sentiment Analysis inspires smarter business in Venture Beat’s coverage of expert.ai’s natural language platform. Then, get started on learning how sentiment analysis can impact your business capabilities.
Read also: Difference between Natural Language Processing and Text Mining and NLP for Big Data: What everyone should know?