The Evolution of AI Emotion and Sentiment Analysis
NLP/Text AnalyticsOpinionsentiment analysisposted by April Miller March 18, 2022 April Miller
Artificial intelligence emotion and sentiment analysis has come a long way over the years and is on track to revolutionize the AIs of the future. Some wonder if it can ever truly understand human emotions, but computer scientists are focusing on training AI to recognize these feelings correctly. The technology has faced challenges over the years but continues to hold valuable potential in developing next-generation AIs.
A Unified Theory of Emotion
The first challenge for AI emotion and sentiment analysis was determining whether feelings can even be categorized. How do we teach an AI to understand them if they can be? As recently as the 1960s, scientists weren’t even in agreement that certain facial expressions were universally connected to specific emotions. Today, this concept is foundational to AI emotion analysis.
Attempts at teaching AI to understand emotion are based on a similar theme: facial expressions indicate how people feel. Algorithms that can identify them can connect them to the right sentiments. AI cannot understand the concept of human emotions, so this is actually just advanced image labeling.
Google’s GoEmotions project is a particularly important step forward in AI emotion and sentiment analysis. GoEmotions combines natural language processing and emotion classification, resulting in a database of 28 unique categories. The project’s data revealed some interesting insights about the emotions humans most often express. For example, Google had the most data samples for “admiration” and the least for “grief.” Ratios like this may help AI understand what feelings are rare or extreme.
Innovation and Business
Growth in AI emotion and sentiment analysis took off when businesses began investing in it, often for hiring purposes. Managers use it to get insights about a candidate based on their facial expressions. This can hint at someone’s honesty and enthusiasm, as well as certain personality traits specific to the role.
For example, a company might want a customer service representative to be friendly and approachable. A good manager might be someone who can be calm and assertive, even in stressful situations like a job interview. Interestingly, some businesses found that applying AI for emotion and sentiment analysis resulted in more diverse new hires. This is believed to be due to a lack of racial and gender bias.
Of course, AI has made headlines for failing to remove these biases in other cases. For example, Amazon faced backlash in 2018 when its application analysis AI showed bias against female candidates. Job applicants may also find it uncomfortable that an AI is assessing their emotions rather than a hiring manager. That’s why the use of AI emotion analysis in hiring and other business-related applications has been slow to take off on a large scale.
Can AI Simulate Emotion?
AI has captured the imaginations of sci-fi fans and leading programmers alike for decades. Scientists continue to pursue ever-more advanced algorithms, setting lofty goals for future professionals and attracting wealthy investors. It’s no wonder that computer science is one of the top majors for college students today. Young people are fascinated by AI and programming, which have become doorways to high-paying jobs.
It seems almost inevitable that, given the influx of new computer science graduates passionate about the industry, we will invent AI that can match humans in terms of intelligence and independent thought one day. The question is: Can AI ever simulate human emotion realistically? AI emotion and sentiment analysis is the first step toward making that possible.
Scientists are creating a massive base of training data for next-gen AI by teaching it to recognize and categorize a range of natural emotional expressions. They will have to train the AI to understand what emotions to use in certain circumstances. They will also need to streamline sentiment analysis to work rapidly. AI can then pull from its known set of emotions and expressions to simulate a natural emotional response.
Samsung is already pursuing human-like AI. In 2020, Samsung’s STAR Labs unveiled a project called “Neon.” The Neon AIs are more than algorithms. They have digital bodies, each with a unique appearance and face. Neons are still in the early stages of their development, but they will one day be able to express emotions like a human and have unscripted conversations in real-time. Mastering AI emotion and sentiment analysis is a key stepping stone toward making technology like this mainstream.
The Future of Emotion
Fiction is riddled with AI robots and beings that are beloved characters. Scientists have been dreaming of making AIs like these a reality for generations. There is something almost innately appealing about creating technology that feels like a friend and can understand and reciprocate feelings. Developments in AI emotion and sentiment analysis are the roots of the far-future AIs that will one day be unrecognizable from today’s algorithms.