AI-Based Sentiment Analysis Improves Customer Experience

March 22, 2021
AI-Based Sentiment Analysis Improves Customer Experience Effective IT Asset Management During COVID-19 | Essentials for Productivity, Security, and Resilience

This is the THIRD blog in our series that unravels the mystery behind the question What is AITSM and Why Do CIOs Need to Face It? In our first blog post, we explained how AI Enables the Transformation of ITSM in 2021 and Beyond. Our second article examined why Companies Bet BIG on Conversational UI and Virtual Agents, as the influx of billions of dollars is fueling dramatic advances in predictive analytics and deep learning. Today, we will explain how AI-based Sentiment Analysis improves customer experience, as companies are accelerating the ability for algorithms to infer preferences from what customers post via various communication channels.

Sentiment analysis is the process of identifying and extracting customer opinions and “feelings” from a segment of text. The analyzed data quantifies the customer’s reaction toward certain products, services, or ideas, and reveals how that person “feels” about those things. Customer sentiment can range anywhere from positive or neutral all the way down to negative. Regardless of where a customer’s sentiment falls within that spectrum, sentiment analysis provides valuable information that helps businesses refine their products, services, brand image, and more.

What is AI-Based Sentiment Analysis?

ChangeGear employs an innovative method for analyzing sentiment, known as Deep Neural Network (DNN), which improves the accuracy of machine learning by analyzing the context between words and plotting it as a measure of distance. Words that are contextually alike have a very short distance between them, while words that are contextually different have a very long distance between them. This forms clusters or groups of words that are very similar, which allows SunView’s ML to build more accurate prediction models.

In the simplest application, Sentiment Analysis will tell you if a person’s opinion is positive, neutral, or negative. On a small scale, this information allows you to tailor your chatbot to respond to a user’s sentiment. On a large scale, you can uncover themes of sentiment to detect how people feel about your product or service. Sentiment Analysis delivers a smarter and more human-like artificial intelligence, which can respond in unique ways based on the emotions that users convey through their text conversations.

What Are Utterances?

Utterances are input from a user that your AI application needs to interpret. An utterance is the smallest unit of continuous speech, beginning and ending with a clear pause. Virtual assistants can process utterances that consist of multiple sentences individually, in a logical order, or simultaneously based on the overall request of the user. Examples of utterances include:

  • Raise the priority of incident IR-0000028 to the next higher level.
  • Can you please raise the priority on my open ticket?
  • Have any company-wide issues been reported today?
  • Is there an outage?
  • How do I setup company email on my cell phone?
  • I need help setting up email on my mobile device.

What Are Intents?

Intents refer to what a user wants to accomplish. Most intents are simple, discrete tasks like “Reset Password,” “Setup Email,” and “Generate Reports,” which are typically described using a verb and noun combination. These types of intents initiate a dialog with the user to capture more information, collect and update data from remote systems, and inform the user of the progress of their request.

The goal of intent recognition is to match a user’s utterance with its intended task or question. SunView Willow AI™ determines a user’s intention using a chatbot model that defines the combinations of words that typically indicate an intent.

What Are Entities and How Do You Extract Them?

Entities are anything that defines, shapes, or modifies the intent of the user. Entities can be dates, times, and locations that are required to carry out the intent. The goal of entity extraction is to identify the elements that are needed to complete the task (i.e., intent). These elements can be simple items like numbers and dates, or they can be more complex items like IP addresses, computer names, and user-defined items such as categories.

Entity extraction is a text analysis technique that uses Natural Language Processing (NLP) to automatically pull-out specific data (e.g., requestor name, incident number, priority) from unstructured text and classify it according to predefined categories. These categories are named entities—the words or phrases that represent a noun. This includes proper names in addition to numerical expressions of time or quantity such as number of hours worked, or project start and end dates.

In the context of your IT organization, entity extraction enables teams to find meaningful information in large amounts of unstructured text data. Sifting through hundreds of emails, customer support tickets, or text chats would take countless hours of manual work. But thanks to automated entity extraction, you can get the exact data that you need in just a few seconds.

How Do You Extract the Right Data from Customer Support Tickets?

As your IT organization grows, keeping up with customer support tickets becomes more challenging. Using an IT Service Management (ITSM) system, like ChangeGear Service Manager, allows you to manage and organize all your tickets in one place. But how do you make sense of them all? Entity extraction tools, like SunView Willow AI™️ Sentiment Analysis, help assesses user satisfaction in real-time, thereby improving your ability to prioritize and escalate tickets.

ChangeGear Service Manager with SunView Willow AI™ has the power to analyze complex sentiments such as, “My desk phone works fine, but I can’t dial out using the softphone.” By recognizing there are two sentiments contained in this quotation, ChangeGear returns a positive sentiment for the desk phone, but a negative one for the softphone. This is critical for drilling down to the granular level necessary for your IT department to understand where it is performing well and where it needs to improve.

Understanding a sentence like “I hate XYZ Corporation’s refund policy on its new smart phone” requires more sophisticated NLP capabilities called entity-based Sentiment Analysis, which recognizes different types of entities that attract sentiment such as people, places, things, organizations, brands, products, and much more. ChangeGear understands that XYZ Corporation is a company and that the refund policy belongs to XYZ Corporation. SunView Software’s technology can also discern that XYZ Corporation is the manufacturer of the smart phone, which is necessary to understand the sentiment that is being expressed in this comment.

Why Do You Need AI-Based Sentiment Analysis?

Customer interactions—whether they indicate positive, neutral, or negative sentiment—can be used to circumvent issues, inform internal teams of problems, and influence new and existing customer behavior. AI-based Sentiment Analysis reveals how individual customers feel about your products, services, and policies. This information can help you identify areas that may need to be improved by:

  • Capturing IT effort that is overlooked or misinterpreted by Key Performance Indicators. KPIs such as call duration are not necessarily the best way to measure the effectiveness your IT support staff. For example, a long phone call may mean that your agent is handling a complex issue—not having trouble resolving it. You can use Sentiment Analysis to identify the agents that are consistently involved in calls with a positive sentiment, so you can reward them and use them to mentor less experienced team members.
  • Providing more data for root cause analysis. By pulling sentiment data into your IT department’s KPI reports, you can find correlations that might otherwise be hidden. You can use line charts, for example, to examine your rate of customer retention plotted versus the number of calls with negative sentiment. Then, you can listen to recordings of the phone calls that have negative sentiment and correlate to a decrease in retention to find out why customers are leaving.
  • Guiding quality assurance auditors to areas that require attention. Auditors do not have enough time to listen in on every phone call and monitor every interaction for quality. Sentiment Analysis helps identify the calls that had negative sentiment, giving auditors a good starting point for their reviews.
  • Amplifying the Voice of Customer. Sentiment Analysis data can be combined with post-call surveys to reveal more information about how customers really feel about their interactions with your agents.

How Can SunView Software Help?

Internal and external customers expect fast and personalized experiences. When they reach out to your service desk, they want to feel like someone is really listening and cares about helping them. But it can be hard to provide a great experience when thousands of tickets are flooding your IT department. With AI-based Sentiment Analysis, you can identify critical or urgent issues as soon as they arrive and assign them a high priority. Then you can set up rules to automatically route urgent tickets to the right team, agent, or chatbot.

SunView Software offers real-time Sentiment Analysis within tickets so agents can instantly gauge the tone of a customer’s response. Powered by SunView Willow AI™️ technology, automated Sentiment Analysis helps customer support teams streamline their ticket workflows and function proactively to resolve customer support issues.

Related Posts