SEMO: A semantic model for emotion recognition in web objects

In this work, we present SEMO, a Semantic Model for EmotionRecognition, which enables users to detect and quantify theemotional load related to basic emotions hidden in short,emotionally rich sentences (e.g. news titles, tweets, captions).The idea of assessing the semantic similarity of concepts bylooking at the occurrences and co-occurrences of termsdescribing them in pages indexed by a search engine can bedirectly extended to emotions, and to the words expressing themin different languages. The emotional content associated to aparticular emotion for a term can thus be estimated using webbasedsimilarity measures, e.g. Confidence, PMI, NGD andPMING, aggregating the distance computed by a model ofemotions, e.g. Ekman, Plutchik and Lovheim. Emotions areranked based on their similarity to the analyzed text, describingeach sentence through a vector of values of emotion load, whichform the Vector Space Model for the chosen emotion model andsimilarity measures. The model is tested comparing experimentalresults to a ground truth in literature. SEMO takes care of boththe phases of data collection and data analysis, to produceknowledge to be used in application domains such as socialrobots, recommender systems, and human-machine interactivesystems.