Currently, there are several automated tools that allow organizations to know the qualitative comments of their customers.
For companies it is very valuable to know the opinion of their customers and stakeholders. Traditionally, the way to know what your thoughts and feelings are towards a certain product, service, company or company policy, has been through surveys.
“The problem is that these surveys cannot capture important emotional responses and, as a result, they end up missing critically important feedback,” explains Mohamed Zaki -Deputy Director of the Cambridge Service Alliance (CSA) at the University of Cambridge-, Janet R. McColl -Kennedy -professor of marketing and director of research at UQ Business School, University of Queensland- and Andy Neely -founder of Cambridge Service Alliance (CSA)- in an investigation published in the Harvard Business Review magazine.
Researchers say there is a “data goldmine” if you know where to look for it and how to analyze it, as customers often reveal their true feelings when they can provide open responses, and open responses are a much more reliable predictor of behaviors.
"This technology aims to determine the feeling or emotion implicit in an opinion," explains John Atkinson, professor and researcher at the Universidad Adolfo Ibáñez, Chile, and PhD in Artificial Intelligence from the University of Edinburgh. Atkinson also stresses the importance of having tools that consider the different subjectivities of language: “The machine takes an opinion and classifies it, but there is a much larger range of options. The first step, before discriminating what the type of opinion is, is to look for messages that talk about the subject. But sometimes you have to infer what is implied and many systems infer without context.
This is where the AI comes into play. “The role of artificial intelligence is in language processing. It is generally assumed that an opinion is a bag of words. But the reality is not like that, there is the way in which they are connected, the subjectivities, ironies, superlatives”, adds Atkinson.
To illustrate it, exemplify it with the phrase "Spiderman is not a bad movie". This sentence indicates a positive comment about the film, but a system without the proper context could recognize the words "no" and "bad", classifying the sentence as a negative comment.
Zaki, McColl-Kennedy and Neely explain that the tools available right now generally categorize feedback into positive and negative, but in their research they defined seven dimensions for mapping customer feedback keywords.
“AI algorithms can capture the specialized vocabulary used by customers and combine their views expressed in their own words with traditional rating scales to obtain insights. These insights can directly shape short- and long-term actions to retain customers.
Criteria to take into account
John Atkinson says that companies that want to use this technology must take a number of considerations into their decision. First of all, they must check if they have information available to feed the system. Then, it is important to analyze whether the chosen solution considers biases and imbalances and which technology would be best applied, depending on the complexity of the messages to be analyzed and the nature of the texts. Finally, to better understand and analyze the information and thus make informed decisions, it is important to discuss how the results are going to be connected with elements that give them significance and validity, as John Atkinson explains.
In 2021, Alaya developed a customized solution for a mining company, with these characteristics, which was fed with keywords taken from community newspapers and social networks, to measure the opinion of employees and community members about a site. Its application was crucial for the relationship strategy of the company with its environment.