Even in the digital context, people express needs and desires at all times: knowing how to read and interpret them, thanks to technologies and skills, is essential for designing better experiences.
Online conversations: blogs, reviews, comments, social networks, nowadays more than 50% of the world population uses them. It’s hard to know someone who hasn’t exposed himself at least once in these contexts. Every day a huge amount of data relating to our thoughts and opinions is poured into the net.
Starting from this spontaneous sharing, is it possible to bring out value for a company?
Through the sentiment analysis activity, which tries to exploit this source of data, by collecting and analyzing texts, it is possible to gather valuable information on people’s needs and desires with respect to products, services and much more.
Natural Language Processing, known as NLP, is that branch of artificial intelligence that makes it possible for a computer to understand the meaning of sentences, so that it can return an output.
“Ok Google, show me the restaurants in the area”, the virtual assistant is an example that is part of our daily lives and is based precisely on its ability to process natural language.
Sentiment analysis is another research context in this area, whose main purpose is to understand the general sentiment that prevails, however, in a written text. Its output can simply express the polarity of the text, “positive”, “negative”, “neutral”, or try to interpret specific feelings such as “anger”, “happiness”, “sadness”, etc., according to the sensitivity and sophistication of the approach used.
There are three main methods to carry out this type of analysis: the detection of keywords, lexical affinity and methods based on machine learning algorithms.
While the former returns an output based on the presence in the text of words that express emotions in an unequivocal way, the latter also attributes an affinity to a certain feeling to words not strictly related to the emotional sphere.
Finally, with the design of algorithms based on machine learning, we try to take into account not only individual words in the analysis, but also their relationship within the sentence, trying to overcome what are currently the biggest obstacles in this area research, i.e. the more ambiguous and articulated aspects of the language, such as the need to know the context of a speech or the use of irony and sarcasm in which positive words can be used to express a negative feeling “How wonderful to go to work knowing that 2 days out of 5 the metro breaks down!”
The potential of sentiment analysis could be exploited in very different contexts: from managing the reputation of important public figures, to the fields of communication and marketing.
Take a product review for example. Giving a judgment through a “vote” is certainly useful for summarizing and providing immediate feedback, but if on the other hand you want to go deeper into this evaluation to understand what went right and what went wrong, sentiment analysis could make the difference.
We could define our latest purchase, a washing machine, giving it, on a scale of 1 to 5, a grade of 4 and comment “It has all the washing programs I wanted, but it makes a terrible noise!”. Here, “4” is almost the maximum assignable value, but in the comment you can find the details of our overall feedback.
Being able to extrapolate opinions and attitudes towards a certain product or service, therefore, could help to obtain information useful for improving them, so that they meet and best respond to people’s needs and all this, by exploiting spontaneous sources of data that are difficult to manage otherwise.
A further demonstration of how listening to experiences, even if expressed within a digital context, can bring immense value to the design of quality experiences, truly tailored to those who will experience them.