Whether celebrating life’s greatest highs or coping with the lows, or just every mood in between, music has outlined my existence from as long as I can remember.

While my parents argue that Walkmans, CDs and portable mp3 players transformed the music world, I believe that today’s online streaming applications like Spotify, have truly revolutionized how music lovers consume music, by reading their minds, and providing them the quick fix they need anytime, anywhere. Soon after Spotify and I, quite literally discovered each other, I was dying to know how it all worked. A few weeks of obsessive research led me to learn about how its Artificial Intelligence (AI) and Natural Language Processing (NLP) capabilities have hyper personalized music.

AI is defined as the capability of machines to imitate intelligent human behavior. AI typically uses the development of algorithms and computer programs to perform tasks that need human intelligence. Various aspects of AI include Natural Language Processing, Machine Learning, Deep Learning, Robotics, Expert Systems, Computer Vision, and Ethics.

Of these, Natural language Processing (NLP) is a huge leap forward in AI technology, possibly because it is getting rid of the communication barrier that has existed between machines and humans.

Here’s how that communication barrier existed.

In the past, computers could only work with structured languages. The language had to be precise and clear-cut. To program a computer to perform a task, you had to give it clear instructions, using only the commands that it understood. The syntax had to be precise as well. NLP is removing the need for being so precise. Instead of us having to learn the computer’s language, computers are now learning ours! Since it removes the communication barrier between humans and computers, the potential for the application of NLP is almost limitless.

So how does NLP work?

NLP models work by finding relationships between the parts of language like the letters, words, and sentences found in a text dataset. 

Natural language processing is hard, mainly because human languages are complex and understanding them needs an understanding of the concepts and the words, and how they’re connected to make sense. For us, it is just regular communication, but everyone knows that words come with a deeper context and when we say something to another person, that person can actually understand what we mean, and all of this communication grows with experience.

So how can we offer that experience to a machine?

The answer is, we need to provide it with a lot of data to help it learn through experience.

The first working step of an NLP system depends on that particular system’s application. For example, voice-based systems, like Google Assistant or Alexa, translate words into text. Usually, this is done using the Hidden Markov Models (HMM) system. The HMM use mathematical models to figure out what a person has said and translate that into text that can be utilized by the NLP system.

The next step is the actual understanding of the context and the language. Though the techniques vary a little bit from one NLP system to another, they mostly follow a similar format. The systems attempt to break every word down into its noun, verb and other parts of speech. This happens through a series of coded rules which depend on certain algorithms. These algorithms integrate statistical machine learning, in order to help figure out the context. For NLP systems other than speech-to-text, the system skips the initial step and directly moves into analyzing the words, utilizing the algorithms and grammar rules.

The final step is the ability to categorize what a person says in different ways. The results get utilized in several ways depending on what the NLP system is actually trying to do.

Since we are talking about how a Natural Language Processing system works, let us look at its key components, Syntactic and Semantic analysis. Syntax stands for the arrangement of words in a sentence, so that they can make grammatical sense. Syntactic analysis is used to assess the way the natural language gets aligned with the grammatical rules. Syntactic and Semantic Analysis differ in the way text is analyzed. In the case of syntactic analysis, the syntax of a sentence is used to interpret a text. In the case of semantic analysis, the overall context of the text is considered during the analysis.

Two popular methods are applied to implement an NLP system – machine learning and statistical interference.

NLP systems have a wide range of applications today, which are constantly growing and transforming as the world transforms.

Chatbots for customer call assistance, language translation programs

for translating languages before a human translator gets involved, sentiment analysis

for analyzing the emotional state, attitude, and mood of people posting messages on social media platforms and descriptive analytics for capturing customer feedback for products or services.

Applications like search autocomplete and search autocorrect are widely used to locate correct search terms, or automatically correct incorrect terms.

Did you know that spell check is also a gift of an NLP application?

My greatest NLP gifts come in the form of my daily Spotify music picks. Spotify’s NLP algorithms constantly scan the web to find blogs, articles or any other text about music, and come up with a profile for each song. With all this scraped data, the NLP algorithm can classify songs based on the kind of language used to describe them and can match them with other songs that are discussed in the same way. Artists and songs are assigned to classifying keywords based on the data, and each term has a sort of “weight” assigned to it. A vector representation of the song is created, and Viola! That is used to suggest similar songs! Of course, these systems rely a lot on user data to curate personalized playlists. Spotify's algorithm analyzes your listening history, favorite genres, and the time of day you listen to music.

In conclusion, NLP is a crucial aspect of AI that is rapidly changing the world, by helping computers understand and generate our natural language.

All this is awesome, but shouldn’t the notion of intelligence include its creative, and wild elements? Surely there are unpredictable elements in our intelligence and language that data-processing algorithms may find elusive?  How likely is it that algorithms will be able to acquire these unpredictive qualities that are needed to solve the problems humanity is facing? And if they can’t, would Artificial Intelligence remain an oxymoron?

I hope not. For now, I’ll enjoy my music!

Bibliography
 

“A Complete Guide to Natural Language Processing,” DeepLearningAI, January 11, 2023.  deeplearning.ai/resources/natural-language-processing/

Sen, Ipshit. “How AI helps Spotify win in the music streaming world,” Outside Insight. outsideinsight.com/insights/how-ai-helps-spotify-win-in-the-music-streaming-world/

Svensson, Jakob. “Artificial intelligence is an oxymoron: The importance of an organic body when facing unknown situations as they unfold in the present moment,” National Library Of Medicine, National Center for Biotechnology Information, November 5, 2021. pubmed.ncbi.nlm.nih.gov/34754144/

Overby, Stephanie. “Artificial intelligence vs. natural language processing: What are the differences?” The Enterprisers Project, February 26, 2020. enterprisersproject.com/article/2020/2/artificial-intelligence-ai-vs-natural-language-processing-nlp-differences

Giattino, Charlie, Mathieu, Edouard, et al. “Artificial Intelligence,” Our World in Data. ourworldindata.org/artificial-intelligence

Miller, Evelyn. “How do Natural Language Processing systems work?” MagniMind Academy. magnimindacademy.com/blog/how-do-natural-language-processing-systems-work/

Kaput, Mike. “How Spotify Uses Artificial Intelligence-and What You Can Learn from It,” Marketing Artificial Intelligence Institute. September 19, 2022. marketingaiinstitute.com/blog/spotify-artificial-intelligence

 Tambekar, Akshad. “How Spotify Uses Machine Learning Models to Recommend You the Music You Like,” Great Learning. August 29, 2022. mygreatlearning.com/blog/how-spotify-uses-machine-learning-models/