Predicting Social Emotions from Readers’ Perspective

Hidden topic–emotion transition model for multi-level social emotion detection

“…Missing someone Approaches for social emotion detection can be categorized into two types: discriminative-model based and topic-model based. By casting emotion detection into a classification problem, many discriminative-model based methods have been proposed [3,4,5], such as the logistic regression model with emotion dependency [6] and the social opinion mining model based on K-Nearest Neighbour (KNN) [7]. However, such approaches are unable to reveal the latent topic information in order to understand how the emotions are evoked.…”

“…In the same vein, a relevant label ranking framework for emotion detection is proposed for predict multiple relevant emotions as well as the ranking of emotional intensity [19,20]. A KNN-like approach called social opinion mining model (SOMM) was proposed in [7] where word embeddings were used to calculate word mover's distance of all the documents via solving an optimizing problem. The average emotion distribution of the k nearest neighbors of a test document was then used for emotion prediction.…”

“…To evaluate the performance of the proposed approach for social emotion detection, we conduct experiments on the News Dataset 2016 (ND16) [7], which was collected from the social channel of Sina between January and December of 2016, consisting of 5,258 news documents with 6 reader emotions (i.e., Touching, Anger, Amusement, Sadness, Shock and Curiosity). The statistics are shown in Table 1.…”

“…The statistics are shown in Table 1. Following the same setup as in [7], the first 3,109 documents in chronological order are used for training and the rest 2,149 are for testing. Preprocessing is conducted on ND16 by first splitting each document into sentences through sentence boundary detection, then removing stop words, numbers, person names, punctuations and words appeared in less than 3 documents.…”

“…SWAT is a supervised approach using an uigram model trained to annotate emotion content. SOMM [7] is one of the optimization approaches which employs word embeddings to calculate word mover's distance of all the documents via solving an optimization problem. ConstraintOp [17] employs [8], SLTM [8] and the Affective Topic Model (ATM) [36].…”