When you hear the term Natural Language Processing (NLP), you might think of robots like Sophia or Ava. The truth is though, NLP has been around for decades — with roots that go all the way back to the 1950s when Alan Turing first wrote his paper on the Turing test.
NLP is so much a part of our lives that we, as consumers, interact with NLP on a regular basis. Probably without even realizing it!
NLP — not to be confused with Neuro-Linguistic Programming from psychotherapy — deals with the use of computers to process informal human languages. A couple of the most famous uses for this would be sentiment analysis and social media monitoring.
But did you know, there are some even more common everyday uses for NLP?
NLP in daily-use applications
So common that it’s become a meme, autocorrect is an application of NLP that can sometimes result in some awkward or hilarious encounters. However, it is more often a godsend — like when you’re writing an article and make a typo, or when you’re using a mobile that’s too small for your fingers.
Your text messaging app or keyboard uses NLP to identify the most likely term that’s closest to your misspelling and automatically changes it to a more accurate one instead.
Besides autocorrect, NLP is also the same technology behind spell check and autocomplete. And when you type “tailor swit” into a Google search, it’s NLP that produces results to information about Taylor Swift, while asking “Did you mean tailor suit?”
NLP makes it possible for companies to create amazing experiences for their customers. Applications of NLP like sentiment analysis, smart search and chatbots are some of the ways that NLP can be used for customer experience (CX) technology.
Sentiment analysis and its applications
As its name suggests, sentiment analysis is about extracting opinions and making sense of them in order to achieve a specific goal. This could be a business goal, an advocacy goal or a public relations goal.
Generally, sentiment analysis algorithms aim to determine whether user reactions are positive, negative or neutral. More specifically, there are four types of sentiment analysis and each has its applications.
The first type is Emotion Detection Sentiment Analysis. This is probably the one we are most familiar with. It involves identifying emotional states by studying text inputs. This method usually uses word lists or symbols that have been segregated based on whether they’re positive or negative, as well as machine learning algorithms. This type of sentiment analysis allows businesses to discover how their customers feel about their products.
The second type is Fine-grained Sentiment Analysis. For this, it could potentially go beyond the simple positive/negative emotion analysis and determine the extent of the polarity of the sentiment as well. For example, it could take into account the fact that five stars in an online product review could be very positive, while 0 stars is very negative, as well as the specifics in between.
The third type is Aspect-based Sentiment Analysis. This can be used to drill-down further into the specifics and discover what aspect of a product or service affects user sentiment. Rather than just classifying a user review, for example, as just positive or negative, aspect-based analysis lets businesses see their users’ perception towards specific parts of their product or workflow.
The fourth type is Intent Analysis. This is typically used to determine the underlying intent within the text input. If a customer says something about a product, is it because they plan to buy it? Or if they have already bought the product, are they planning to use it? How are they planning to use it? It’s a way for companies to understand their users better and plan the marketing or product development in a more targeted manner.
Increasing use of sentiment analysis
It’s a standard these days, for businesses to conduct some kind of sentiment analysis and it’s vital to have enough accurate sample datasets that can be used for training machine learning algorithms.
NLP is a common part of the CX process and many businesses are making use of the currently available datasets and methods to train their models. However, most of the available datasets are for the English language and are focused on Western consumers.
Using them in Southeast Asia would be pointless as not only do these datasets not take into account local languages, they also don’t include the nuances of English or emoji usage in this region.
Companies looking to expand into resource-poor countries like Thailand, Vietnam and Indonesia — where markets are generally untapped but sizable — might need to look into alternative strategies to conduct sentiment analysis, while keeping costs affordable.