Sense and Sentimentality

The written word is a very powerful form of communication. It gives us the power to translate our own thoughts into a sharable product, to communicate important information to the masses, and to persuade someone to do something or not to do it. For thousands of years, humans wrote down their communications, starting with clay for tablets and reeds for styluses , then with pen and paper, printing presses, typewriters, and finally to computers and smartphones. In fact, it was the invention of computers that sparked the initial interest in large scale analysis of written text. Trying to analyze patterns like words with machines (like computers) is both a new fangled idea and an old-fashioned one at the same time.

NLP stands for Natural Language Processing.

Alan Turing introduced many to NLP and the Turing Test in his 1950 paper where he argued that we can consider computers intelligent if they carry on conversations with humans without humans realizing they are communication with a machine. Even though terms like machine learning and natural language process (NLP) are new ones in the mainstream vernacular, words within them like machine and processing also seem antiquated and perhaps conjure up images of mainframe computers and punch cards.

Score Sentimentality in Power BI

When we read a written opinion, our brains subconsciously run their own algorithms behind the scenes to determine whether the piece has a positive or negative sentiment associated with it. We can also use computer algorithms to determine this sentiment for us. In this week's Power BI Weekly series video, I show how to use the score sentiment function to run this text analytics algorithm through Azure Cognitive Services directly in Power BI. Note this functionality requires a Power BI Premium account, but this includes the per user option!

URL: https://www.dhirubhai.net/learning/power-bi-weekly/calculating-score-sentiment-with-text-analytics-functions

The inspiration for the data I used to illustrate the power of this NLP algorithm came from some articles I recently read where people visit US National Parks (some of the most beautiful places in the world) and return home to give them one-star Yelp reviews for some of the most bizarre and unbelievable reasons.

I decided to create my own fictional reviews for the fictional Landon Hotel in London. If you watch the video, you'll see I created reviews for customers that are either impossible to keep happy or gave so many compliments it almost felt cringy. I then ran the paragraphs of text through the Power BI score sentiment algorithm to see what score (between 0 and 1) the results returned.

Other Text Analytics Algorithms in Power BI

There are more built-in text analytics options for NLP to also try out in Power BI (also only with Premium accounts). This includes algorithms like language detection and key phrase extraction that run through Azure Cognitive Services directly in Power BI, much like the score sentiment algorithm does. You can check out how to run these two other text algorithms in the video below from my Power BI: Integrating AI and Machine Learning course I created for the LinkedIn Learning library last year.

URL: https://www.linkedinom/learning/power-bi-integrating-ai-and-machine-learning/utilizing-text-analytics-algorithms

GPT Models

Lastly, we can't talk about the future of algorithms like NLP without mentioning the unbelievable potential power of GPT models. In 2019, the GPT-2 (Generative Pretrained Transformer 2) NLP algorithm wrote an entire story about talking unicorns living in the Andes based on only two sentences of input text. This?OpenAI ?language model ran ten times using a lot of computing power and a neural network architecture to generate written prose. This text looked so much like humans actually wrote it that the algorithm creators decided it was too dangerous to release its code for fear of what it could potentially do, especially on a large scale (like generating massive volumes of fabricated news stories for example).

Up next week: the Python visual in Power BI! As I mentioned in last week's newsletter, Python is taking over the world, but I think its popularity is a great thing. It also means that Power BI directly supports building Python visuals through the standard visuals.

-HW