Adaptive Recursive Neural Network for Target-dependent Twitter Sentiment Classification

By Li Dong, Furu Wei, Chuanqi Tan, Duyu Tang, M. Zhou, Ke Xu. 2014

984 citations recorded by the source corpus.

We propose Adaptive Recursive Neural Network (AdaRNN) for target-dependent Twitter sentiment classification. AdaRNN adaptively propagates the sentiments of words to target depending on the context and syntactic relationships between them. It consists of more than one composition functions, and we model the adaptive sentiment propagations as distributions over these composition functions. The experimental studies illustrate that AdaRNN improves the baseline methods. Furthermore, we introduce a manually annotated dataset for target-dependent Twitter sentiment analysis.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Erik Cambria
2. Lidong Bing
3. Mohit Bansal
4. Bing Liu
5. Jianfeng Gao
6. B. Liu
7. Sebastian Ruder
8. Minlie Huang

Citations in the corpus

• Attention-based LSTM for Aspect-level Sentiment Classification (2016, methodology): “Classical models including Recursive Neural Network (Socher et al., 2011; Dong et al., 2014; Qian et al., 2015), Recursive Neural Tensor Network (Socher et al.”
• Syntax-Aware Aspect Level Sentiment Classification with Graph Attention Networks (2019, background): “Some early work also tries to leverage the syntax structure (Dong et al., 2014; Nguyen and Shirai, 2015).”
• Deep Learning--based Text Classification (2020, background): “Some of the most popular datasets include SemEval-2014 Task 4 [161], Twitter [162], SentiHood [163], to name a few.”
• Interpreting Neural Networks to Improve Politeness Comprehension (2016, methodology): “Recent work has also applied neural network based models to sentiment analysis tasks (Chen et al., 2011; Socher et al., 2013; Moraes et al., 2013; Dong et al., 2014; dos Santos and Gatti, 2014; Kalchbrenner et al., 2014).”
• Target-Dependent Sentiment Classification With BERT (2019, methodology): “2https://storage.googleapis.com/bert_models/2018_10_18/uncased _L-12_H-768_A-12.zip 3https://github.com/google-research/bert Rec-NN [6] first applies rules to transform the dependency tree of a sentence and puts the target at the tree root, and then learns the semantic composition of the sentence vi”
• Multi-Task Learning of Pairwise Sequence Classification Tasks over Disparate Label Spaces (2018, positive): “We use the data from Dong et al. (2014).”
• Interactive Attention Networks for Aspect-Level Sentiment Classification (2017, background): “…have designed effective neural networks to automatically generate useful low-dimensional representations from targets and their contexts and obtain a promising result on the aspect-level sentiment classification task [Dong et al. , 2014; Vo and Zhang, 2015; Wang et al. , 2016; Tang et al. , 2016b].”
• Gated Neural Networks for Targeted Sentiment Analysis (2016, methodology): “The method of Vo and Zhang (2015) avoids the reliance on automatic syntactic parsers, which can be highly inaccurate on tweets, thereby leading to improved sentiment classification results compared with Dong et al. (2014).”
• Deep learning for sentiment analysis: A survey (2018, methodology): “Considering the limitation of work by Dong et al. 68 and Vo and Zhang 69 , Zhang et al. 72 proposed a sentence level neural model to address the weakness of pooling functions, which do not explicitly model tweet-level semantics.”
• Aspect-based Sentiment Classification with Aspect-specific Graph Convolutional Networks (2019, methodology): “Tai et al. (2015) showed that LSTM with dependency trees or constituency trees outperformed CNNs. Dong et al. (2014) presented an adaptive recursive neural network using dependency trees, which achieved competitive results compared with strong baselines.”
• Transformation Networks for Target-Oriented Sentiment Classification (2018, methodology): “• AdaRNN (Dong et al., 2014): It learns the sentence representation toward target for sentiment prediction via semantic composition over dependency tree;”
• Multi-grained Attention Network for Aspect-Level Sentiment Classification (2018, background): “Recursive Neural Network (Socher et al., 2011; Dong et al., 2014; Nguyen and Shirai, 2015) are studied to conduct semantic compositions on tree structures, and generate representations for prediction.”
• Like It or Not (2016, background): “Dong et al. [2014] proposed an Adaptive Recursive Neural Network (AdaRNN) for entity-level TSA.”
• Transfer Capsule Network for Aspect Level Sentiment Classification (2019, background): “In recent years, a number of deep learning methods have been proposed for aspect-level sentiment classification (ASC) tasks (Dong et al., 2014; Vo and Zhang, 2015; Tang et al., 2016; Wang et al., 2016a; Ma et al., 2017; Li et al., 2018; Ma et al., 2018; Wang et al., 2018a).”
• Target-Dependent Sentiment Classification with Long Short Term Memory (2015, methodology): “Since the benchmark dataset from [Dong et al., 2014] comes from Twitter, we compare between sentiment-specific word embedding (SSWE) 2 [Tang et al., 2014b] and Glove vectors 3 [Pennington et al., 2014] .”