Sentiment measurement is a well-known application of Natural Language Processing in Artificial Intelligence and has been in slow development for decades. It is essentially meant for machines to better understand humans, through understanding their language.
There are different approaches to the problem. Some rely heavily on machine learning. Some are domain-specific, while others attempt to support open-domain opinion mining (i.e. to find the opinions of a public or actor in a wide range of discursive contexts).
The primary emotions
InterTextueels Dutch emotion recognition tool provides a much more fine-grained analysis than traditional sentiment analysis. While existing sentiment analysis solutions classify texts as positive, negative or neutral, our emotion detection service reveals the presence of one or more of the primary and universal emotions: happiness, sadness, surprise, fear, disgust and anger.
Emotion detection through grammar analysis
Our emotion extraction method involves a knowledge-based approach and fully parsing the grammar of sentences to allow a flexible definition of when an emotion is expressed inside a text. We apply additional techniques to weed out false positives.
We aim to support all possible text domains: ranging from high-brow literature to journalism to online gamers slang. We have already tested our emotion recognition software in many different contexts and welcome new challanges posed to us.
Customers can supply us their text corpus and we will be able to manually test, improve and fit our product to their dataset. We apply modern machine learning techniques to improve accuracy and completeness. We are even experimenting with irony detection. In our research blog, we will provide working examples of the emotion recognition service. Please look at the demonstration pages as well, to see our service in action in the domain of social media like Twitter.
- Next s