Sentiment Analysis with Feeling

September 25, 2020

As AI technology progresses, so too does the field of sentiment analysis. What could go wrong? Sinapticas explores “How Algorithms Discern Our Mood from What We Write Online.” Reporter Dana Mackenzie begins with an example we can truly relate to right now:

“Many people have declared 2020 the worst year ever. While such a description may seem hopelessly subjective, according to one measure, it’s true. That yardstick is the Hedonometer, a computerized way of assessing both our happiness and our despair. It runs day in and day out on computers at the University of Vermont (UVM), where it scrapes some 50 million tweets per day off Twitter and then gives a quick-and-dirty read of the public’s mood. According to the Hedonometer, 2020 has been by far the most horrible year since it began keeping track in 2008. The Hedonometer is a relatively recent incarnation of a task computer scientists have been working on for more than 50 years: using computers to assess words’ emotional tone. To build the Hedonometer, UVM computer scientist Chris Danforth had to teach a machine to understand the emotions behind those tweets — no human could possibly read them all. This process, called sentiment analysis, has made major advances in recent years and is finding more and more uses.”

The accompanying “Average Happiness for Twitter” graph is worth a gander, and well illustrates the concept (and the ride that has been 2020 thus far). The article is a good introduction to sentiment analysis. It contrasts lexicon-based with the more complex neural network approaches. We learn neural networks may never completely eclipse lexicon-based systems because of the immense computing power required for the latter. Hedonometer, for example, uses a lexicon.

Mackenzie also describes several applications of sentiment analysis, like predicting mental health, assessing prevailing attitudes on issues of the day, and, of course, supplying business intelligence. And the hedonism of the hedonometer, of course.

Cynthia Murrell, September 25, 2020

Comments

Got something to say?





  • Archives

  • Recent Posts

  • Meta