Sentiment Analysis and Its Impact To Business
“Everything must be sacrificed, if necessary, for that one sentiment: universality”, said Swami Vivekananda. Sentiment by itself means ‘a view or an opinion that is held or expressed’ and here in his words ‘Universality’ becomes a sentiment, that which can be interpreted differently in a multitude of intellectual or physical conditions. However, the grounded gist of his intent and the clarity of that message has and will remain the same even with the change in timelines.
Simply, such quotes for any reader ingrains deep positive intent and can germinate positivity instantly. We are constantly developing a 21st Century world where skills such as digital literacy, problem-solving, critical thinking, business and economy, peace, cooperation, etc. are being advocated through education and learning, and the foundations for these are flourishing through disrupting technologies such as AI, NL, ML, IOT, DL, etc. One such technology that is changing the way business is being conducted is Sentiment Analysis. For starters, a simple search on the web will define sentiment analysis as a process of computationally identifying and categorizing opinions expressed in a piece of text, especially to determine whether the writer’s attitude towards that topic, product, etc. is positive, negative, or neutral. Traditionally and to this day, companies rely on customer feedback, reviews, user case studies, etc., to determine the present and future health of their product, service, and people. However, this process of collaborating and collating extensive data is laborious, time-consuming and expensive, as most of this comes from text data like emails, support tickets, chats, social media, surveys, articles, blogs, and documents. Enterprises lose billions of dollars in revenue just in trying to streamline the process. Would it not be amazing if companies were able to capture and understand the essence or the consensus of its potential customers before their product, idea or service is even launched just by analyzing historic review data and/or feedback from related domains or sectors? Sentiment Analysis is intended just to do that! Let’s delve a bit deeper.
Sentiment Analysis also is known as Emotion AI or Opinion Mining, is a subject under Natural language processing and a technique under ML and DL. A framework is built to identify, capture, and extract opinions and attributes within the text. The message is analyzed to predict the sentiment or polarity of the text as positive, negative, or neutral. Then the characteristics of the reviewer can be determined to gauge if the entity is a potential customer or a detractor along with the topic that is being highlighted, which needs to be addressed. Since publicly and privately available information over the web and internet is constantly growing, a large number of texts expressing opinions are available on review sites, forums, blogs, and social media. With the help of sentiment analysis systems, enterprises can use this unstructured information and transform it into structured silos of data of public opinions about products, services, brands, politics, or any topic that people can express opinions about. This data can then be used to enhance and develop commercial applications/tools such as marketing analytics, public relations, social media, product feedback, the voice of the customer, product analytics, brand monitoring, customer service, and workforce analytics. Enterprises can also utilize sentiment analysis to make sense of vast sets of unstructured text data by automating business processes, getting actionable insights, and saving hours of manual data processing, in other words, by making teams more efficient.
Sentiment Analysis [SA] by itself, is a vast subject. There are several methodologies being researched and understudy but by a broad classification we can orient all approaches into three prominent categories -a ‘lexical based approach’ that uses language semantics, a ‘rule-based approach’ that collates and categorizes opinions and more commonly a ‘machine learning & NLP method that uses algorithms to determine an emotion. There are systems that focus on polarity (positive, negative, neutral) to systems that detect feelings and emotions (angry, happy, sad, etc.) or identify intentions (interested v. not interested). Let us explore a few.
- Fine-Grain Sentiment Analysis – We can identify the polarity of a review or an opinion and can classify them as positive, negative, or neutral but to make the icing on the cake more attractive and drool-worthy, it would be pragmatic to go far beyond just sentiment and equip SA with the ability to identify polarities and its prevailing emotions as well. (e.g., satisfaction, happiness, excitement, anger, sadness, and anxiety). Data, thus obtained, will deliver accurate, actionable results, thus enabling organizations to implement corrective measures strategically. For example, some words that would typically express frustration like ‘Insane’ (e.g., your product is insane, it chars all my recipes) might also express happiness (e.g., in texts like ‘your product is insane, the steak is well done each time’). We can diagnose that the first review can be assigned as a ‘very negative’ sentiment with ‘frustration’ as the attached emotion and the next review can be assigned as a ‘very positive’ sentiment with ‘satisfaction’ as the attached emotion.
- Feature/Aspect Sentiment Analysis – Instead of classifying the overall sentiment of a text into polarities, feature-based SA allows us to associate specific sentiments with different aspects of a product or service. The analysis then denotes the attributes or components of a product or service and allocates a sentiment to each text by addressing the attached emotion with meaningful insights. This technique can help businesses become customer-centric and place their customers at the epicenter of everything they do. For example, a software company might want to understand the specific sentiments towards different aspects of its product. A review might say: “support was great, but the application GUI is confusing,” which contains a positive sentiment towards ‘aspect customer support’ but a negative sentiment towards ‘aspect user experience.’ A sentiment analysis model might classify the overall sentiment as negative and ignore the fact that the staff did a good job, or vice versa whereas an aspect-based analysis model would differentiate between aspects and allocate a sentiment to each one. Customers want to feel like they’re being listened to, and by using deeper machine learning models like aspect-based sentiment analysis, businesses can send quick, efficient, and personalized responses. And for customer support teams, it means streamlining processes and gaining more valuable insights. Thus, the support team gets a pat on the back, and the development team will need to go back to the drawing board.
- Intent Analysis – There is a need to understand what intentions are behind sentiment and the plausible actions possible for a textual review or a query. Intents are the intentions of the end-user; these intentions or intents are conveyed by the user. We could classify these intents in two categories – casual intents or business intents. It can lead to many benefits for the company and its consumers. It can also help us understand the feedback from the customers about any products, or we can also understand the nature of the service which we have to provide or improve. If we talk about improving products or services, then the intent analysis will provide an upper hand. For example, a chat exchange between a subscriber and a customer executive can clearly underline the need for an intent analysis. ‘Hello, I need assistance in accessing my profile to update my home address.’ The intent here is casual, and the subscriber can be led accordingly. ‘Hello, I need assistance in accessing my profile to update my home address as I am about to place a new order.’ Here the intent is a casual and potential business. The subscriber can then be led to change the address in addition to being assisted in choosing the best available version of the product for the right price. However, in both these scenarios it is easier for a human to decipher the intent as a request, that is casual, or business but machines such as a chat boot can have some problems to identify such intentions. Sometimes, the intended action can be inferred from the text, but sometimes, inferring it requires some contextual knowledge and training as well. Infusing NLP based intent analysis on virtual assistants and automated smart bots can help achieve companies to provide a more human touch.
A globally renowned cab aggregator is systematically integrating its social media engagements to set a new course and use social listening to become more ‘customer obsessed’ as quoted by their head for digital media. For instance, the inability to tip drivers has been a big detractor for the brand and has spurred a lot of anger and disgust. Therefore, a few weeks ago, the company launched 180 Days of Change: a commitment to improving the driver experience that includes the ability to tip. Even though 27% of U.S. sentiment around the tipping announcement was still negative, based on real-time social media data, 56% was neutral, and 17% was positive. What’s more, 35% of the emotion around the announcement indicated anticipation, versus 16% that indicated disgust. 11% also indicated joy, according to the presentation. “Having a breakdown of those emotions is helpful as opposed to the binary positive or negative analysis,” he said.
Then there is one of the biggest e-commerce giants that use SA systems to develop, strengthen and continuously improve their market status quo based on reviews and feedback received not only by them but also their potential competition. SA technology was used internally to measure success rates of existing third-party delivery partners on parameters such as delivery times, delivery demographics, delivery trust, satisfaction indexes, etc. before the company launched their team and department that ensured more than 90% product/item was delivered satisfactorily. By reducing dependency on delivery partners, they could deliver better customer experience.
Sentiment analysis is broader and deeper than the traditional customer survey that asks a participant to assign a value to a statement, such as “strongly agree,” “strongly disagree” or something in between. You can apply sentiment analysis to any unstructured data, such as the text of an email, a Facebook comment or a Tweet. Through machine learning algorithms, sentiment analysis becomes more exact. There is a plethora of companies that are evolving SA systems to cater to every type of business.
We can download applications that are simple lexicon based API’s and provide us with an insight of our review, post or a tweet that we intend to publish harmlessly or otherwise and then there are serious contenders who can predict if your enterprise is built on customer-centric models and principles or the management needs to revisit the ideation chamber again. Just like another famous quote of Swami Vivekananda, “We must go far beyond sentiment when we want to judge.” That is where we believe the future lies.