Thursday 27 March 2025
For years, scientists have been trying to crack the code of understanding what people really mean when they write online reviews. From scathing critiques of products to glowing endorsements, these written opinions can be a powerful force in shaping consumer behavior. But until now, computers haven’t been very good at figuring out what makes them tick.
That’s because online reviews are full of subtleties that humans take for granted – phrases that convey tone and emotion, nuanced language that hints at complex ideas. Computers, on the other hand, rely on straightforward rules to make sense of text. And when it comes to understanding human emotions and intentions, these simple algorithms fall woefully short.
Recently, a team of researchers set out to change this by developing a new type of computer program designed specifically to understand online reviews. The goal was ambitious: create a machine that could accurately identify not just the sentiment of a review (positive or negative), but also its underlying meaning and purpose.
The solution came in the form of a typology – a system for categorizing online reviews into different types based on their content, tone, and purpose. By training the computer to recognize these patterns, the researchers hoped to create a program that could accurately identify the intentions behind each review.
To test their theory, the team developed a dataset of over 1,000 online reviews from various products and categories. They then trained their machine learning algorithm on this data, using it to categorize each review into one or more of the 24 different types they had identified.
The results were impressive: the computer was able to accurately identify the intentions behind each review with an accuracy rate of over 75%. But what’s really remarkable is that the program didn’t just rely on simple keywords or phrases – it actually understood the subtleties of human language, recognizing things like sarcasm and irony that are often lost in translation.
The implications of this technology are far-reaching. For consumers, it could mean getting more accurate recommendations from online retailers, based not just on product features but also on how others have used them. For businesses, it could provide valuable insights into customer sentiment and preferences, allowing them to tailor their marketing strategies and improve customer satisfaction.
But perhaps the most exciting potential application is in the field of artificial intelligence itself.
Cite this article: “Coding Human Intent”, The Science Archive, 2025.
Online Reviews, Sentiment Analysis, Machine Learning, Natural Language Processing, Computer Program, Typology, Categorization, Dataset, Algorithm, Accuracy Rate, Sarcasm, Irony, Artificial Intelligence, Customer Behavior, Marketing Strategies, Consumer Behavior, Product Features,
Reference: Ori Shapira, Yuval Pinter, “Information Types in Product Reviews” (2025).
