18 Sentiment Analysis

Kangaroos on Western Sydney University campus © 2022  Western Sydney University taken by  Sally Tsoutas Western Sydney University Photographer is licensed under an  Attribution-NonCommercial-NoDerivatives 4.0 International

What is sentiment analysis?

Sentiment analysis in marketing is the process of determining people’s opinions about a product or a company by analysing public’s comments on various online platforms. Besides leaving comments on an organisation’s own website, online users may also provide feedback and engage in discussions, and conversations on a variety of different platforms. Examples of such platforms include social media websites (e.g., Instagram), knowledge platforms (such as Quora), media sharing platforms (e.g., YouTube, Vimeo, Spotify), service-providing platforms (e.g., Uber, Airbnb), and several shopping websites (e.g., Amazon). In the olden times, this concept was known as ‘product reviews’. However, now with the amount of information being generated, analysts are able to ‘crawl’ the web and mine relevant data.  Online statements or customer comments are turned into categorical data (like “positive”, “negative” or “neutral”), and summarised to give a manager a bird’s eye view of how the general public is responding to their brand or product.

While sentiments may also be revealed in customer conversations with an organisation’s call center, much of the research in this area has focused on analysing people’s written comments online.

Advantages in Using Sentiment Analysis

Gauging public sentiment has never been easier, quicker, cheaper, and less biased. Sophisticated tools have made sentiment analysis accessible to many businesses. Marketers are able to make swift changes to their campaigns (e.g., changing the music being played in an ad by Expedia) in view of the ‘sentiment’ being expressed. In comparison, using traditional tools of research would be prohibitively expensive and time-consuming. Moreover, the results would also be prone to a degree of human error.

Sentiment analysis can help reveal ‘influencers’ for a brand or a product. While many companies are able to ‘recruit’ social media influencers to market their brands, it is quite possible that a brand already has a cohort of loyal supporters. These are the people who are a brand’s advocates and are able to influence others through their comments.  While social media influencers are often viewed as people with a huge following, there is now a realisation that ‘micro-influencers’ are also important. The number of followers for micro-influencers may be relatively small, yet their bond with their followers is strong. This could be critical when people are seeking sincere advice on matters which are important to them.

Similar to the above point, it is equally important to be able to identify hate speech online, especially when it involves the brand or any of its key values. Brand hate is different from other negative sentiments such as mere disappointment, dissatisfaction, or frustration.  There could be different levels of brand hate with the most extreme one leading to brand bullying.  People who display such a strong negative sentiment are also likely to indulge in negative online word of mouth, online public complaining, and driving boycott campaigns. Sentiment analysis is useful in highlighting such a trend, even if it is only limited to a certain group. Strategies to navigate through such consumer negativity in the digital world is critical.

Steps in sentiment analysis

Below is a simple process through which sentiment analysis could be undertaken:

1. Detecting sentiment
   This is the most basic level of conducting sentiment analysis. It simply provides a quick overview of the overall opinion of customers. This means growing through online comments and extracting opinionated data (such as “I love this!”).  In comparison, there could be other types of data such as objective data (like “the restaurant is located downtown”), which may not display any consumer sentiment at all

 2. Categorising sentiment and identifying the intensity
This step involves detecting whether the sentiment is positive, negative, or neutral. Depending on the type of software being used, managers may also add weighting to these categories, e.g.,

• very positive
• somewhat positive
• neutral
• somewhat negative
• very negative

These categories demonstrate the intensity of the sentiment

3.  Mixed Connotation
Sometimes, the text contains mixed or ambivalent opinions, for example, “staff was very friendly but we waited too long to be served”. One would need to separately  interpret such statements which might be difficult for a simple machine-based program to code and analyse

Source: Fun Robotics ^[1]

Limitations of sentiment analysis

Below are a few, key limitations of sentiment analysis tools as recognized in the industry:

• Machine-dependent sentiment analysis is based on the way people use language to express their opinions. A word’s meaning in the dictionary could be very different from the way people use it in everyday conversations. Sentiment analysis may run into issues when online users utter phrases to display sarcasm
• It is still felt that sentiment analysis is most effective when it is used with large and numerous data sets. Small businesses may find that there is not much data available for their products/services which can be effectively analysed
• Moreover, sentiment analysis is not a one-off activity. It has to be integrated into a firm’s information-gathering strategy. To get real value out of sentiment analysis tools, one needs to analyse large quantities of textual data on a regular basis

Table 2: Examples of different types of emotions, classified into positive and negative sentiments