Home  ::  Ontology Structure  ::  Adjective Noun Pairs  ::  Applications  ::  Downloads  ::  About
We present a novel approach towards sentiment analysis based on the semantic understanding of visual content. To reseach this goal, we automatically construct a large-scale ontology of semantic concepts, each selected based on the following criteria: (1) reflect a strong sentiment, (2) have a link to an emotion, (3) be frequently used and (4) has reasonable detection accuracy.

The construction procedure is illustrated on the right: the process is founded on psychological research such as Plutchik's Wheel of Emotions, a well established psychological model of emotions. First for each of the given 24 emotions defined in Plutchik's theory we retrieve images and videos from Flickr and YouTube respectively to extract their distinct tags (e.g., joy leads to happy, beautiful, and flower).

These tags are then analyzed used to form adjective noun combinations or adjective noun pairs (ANP) such as beautiful flower reflecting a strong positive sentiment or bloody zombie reflecting a strong negative sentiment. Those ANPs are then ranked by their frequency on Flickr and sampled to form a broad and comprehensive ontology containing more than 3,000 ANP concepts.

Tao Chen, Damian Borth, Trevor Darrell, Shih-Fu Chang
DeepSentiBank: Visual Sentiment Concept Classification with Deep Convolutional Neural Networks
arXiv preprint, arXiv:1410.8586, Oct. 2014 -  [paper]

Tao Chen, Felix X. Yu, Jiawei Chen, Yin Cui, Yan-Ying Chen, Shih-Fu Chang
Object-Based Visual Sentiment Concept Analysis and Application
ACM Int. Conference on Multimedia (ACM MM), Orlando, Nov. 2014 -  [paper]

Damian Borth; Tao Chen; Rong-Rong Ji; Shih-Fu Chang
SentiBank: Large-Scale Ontology and Classifiers for Detecting Sentiment and Emotions in Visual Content
ACM Int. Conference on Multimedia (ACM MM), Demo Track, Oct. 2013 -  [this website]

Damian Borth, Rongrong Ji, Tao Chen, Thomas Breuel and Shih-Fu Chang
Large-scale Visual Sentiment Ontology and Detectors Using Adjective Noun Pairs
ACM Int. Conference on Multimedia (ACM MM), Barcelona, Oct. 2013 -  [paper]
Shih-Fu Chang
High-level Semantic Modeling
June, 2014, CVPR Tutorial, IEEE CVPR -  [slides]

Damian Borth
Large-scale Visual Sentiment Ontology and Detectors Using Adjective Noun Pairs
Nov, 2013, Oral Presentation, ACM MM -  [slides]

Shih-Fu Chang
Large-Scale Concepts and Classifiers for Describing Visual Sentiment in Social Multimedia
June, 2013, Keynote Talk, Workshop on Big Data Computer Vision, IEEE CVPR -  [slides]

Damian Borth
tel: +49 631 20575 4184
email: damian.borth [at] dfki.de
Tao Chen
email: taochen [at] ee.columbia.edu
Rongrong Ji
email: rrji [at] ee.columbia.edu
Shih-Fu Chang
email: sfchang [at] ee.columbia.edu