Welcome
|
Understanding visual content by sentiment Visual content such as images and videos does not only contain objects, locations and actions but also cues about affect, emotion and sentiment. Such information is very useful to understand visual content beyond semantic concept presence thus making it more explainable to the user. Click for more details about the ontology. |
Visual Sentiment Ontology & Dataset The goal of this work was to design an ontology of semantic concepts which (1) have a link to an emotion (2) have a strong sentiment and (3) are frequently used on online platforms like Flickr or YouTube. Currently the visual sentiment ontology includes more than 3,000 concepts, with each concept being composed of an adjective and noun e.g. "beautiful sky" or "sad eyes". A list of all concepts can be found under adjective noun pairs list. |
Application Scenarios for the Ontology Once being able to understand the sentiment of images and videos visually different application scenario might the implemented. The output of concept detection systems covering the ontology could be used for example to perform visual sentiment analysis on Twitter images, advertising placement on YouTube or opinion mining on top of different media channels. |
