Sentiment Classification User Modeling with Neural Network for - PowerPoint PPT Presentation

User Behavior Analysis for Sentiment Classification  User Modeling with Neural Network for Review Rating Prediction. Duyu Tang, Bing Qin, Ting Liu. IJCAI 2015, full paper.  Learning Semantic Representations of Users and Products for Document Level Sentiment Classification. Duyu Tang, Bing Qin, Ting Liu. ACL 2015, full paper. 1

Sentiment Classification • Given a piece of text, the task aims to determine its polarity as • Positive / Negative • 1-5 stars • The task can be at • Word/phrase level, sentence level, document level • We target at document-level sentiment classification in this work 2

Standard Supervised Learning Pipeline Training Data Feature Representation Learning Algorithm Sentiment Classifier 3

Feature Learning Pipeline Training Data Learn text representation/feature from data! Feature Representation Learning Algorithm Sentiment Classifier 4

Deep Learning Pipeline Training Data Semantic Composition Feature Representation Word Representation w 1 w 2 …… Learning w n−1 w n Words Algorithm Sentiment Classifier 5

User Behavior is important for SA • From a sentiment analysis perspective , users have different habits to • Assign ratings on IMDB, Yelp … • Use diverse sentiment words to express one’s sentiment Rating Histories Frequently used words good, ok, just soso, User 1 disgusting… awesome, amazing, User 2 excellent, not bad… 6

The Approach • Text semantics gold rating = 2 Softmax v d Pooling Convolution Tanh h 1 h 2 h n Linear …… Lookup w 1 w 2 w n w i : word 7

The Approach • Text semantics • +User-Text Associations gold rating = 2 Softmax v d The model in IJCAI 2015 Pooling Convolution Tanh h 1 h 2 h n Linear × × …… × Lookup U k w 1 U k w 2 U k w n U k : user w i : word 8

The Approach • Text semantics • +User-Text Associations • +User-Sentiment Associations gold rating = 2 Softmax u k v d Pooling Convolution Tanh h 1 h 2 h n Linear × × …… × Lookup U k w 1 U k w 2 U k w n U k : user w i : word 9

The Approach • Text semantics • +User-Text Associations +Product-Text Associations • +User-Sentiment Associations +Product-Sentiment gold rating = 2 Softmax u k v d p j The model in ACL 2015 Pooling Convolution Tanh h 1 h 2 h n Linear × × × × …… × × Lookup U k w 1 P j w 1 U k w 2 P j w 2 U k w n P j w n P j : product U k : user w i : word 10

Experiments • We conduct supervised learning on three datasets • Some details • We build the datasets by ourselves. • Split each corpus into training, development and testing sets with a 80/10/10 split • IMBD is from Diao et al. 2014 • Yelp 2013 & 2014 come from Yelp Dataset Challenge 11

Experimental Results • On IMDB dataset Our Approach 12

Experimental Results • On Yelp 2014 dataset Our Approach 13

Experimental Results • On Yelp 2013 dataset Our Approach 14

In Summary • We encode users (and products) in semantic vector space, and apply them to sentiment classification. • User and product representations can improve the sentiment classification accuracy. 15

Thanks! codes and resources will be publicly available at: http://ir.hit.edu.cn/~dytang 16