Multichannel Variable-Size Convolution for Sentence Classification

By Wenpeng Yin, Hinrich Schütze. 2015

162 citations recorded by the source corpus.

We propose MVCNN, a convolution neural network (CNN) architecture for sentence classification. It (i) combines diverse versions of pretrained word embeddings and (ii) extracts features of multigranular phrases with variable-size convolution filters. We also show that pretraining MVCNN is critical for good performance. MVCNN achieves state-of-the-art performance on four tasks: on small-scale binary, small-scale multi-class and largescale Twitter sentiment prediction and on subjectivity classification.

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

Citations in the corpus

Text Classification Improved by Integrating Bidirectional LSTM with Two-dimensional Max Pooling (2016, methodology): “…2015) 51.4 87.9 - 96.0 - - Molding-CNN (Lei et al., 2015) 51.2 88.6 - - - - CNN-Ana (Zhang and Wallace, 2015) 45.98 85.45 93.66 91.37 81.02 - MVCNN (Yin and Sch¨utze, 2016) 49.6 89.4 93.9 - - - RNN RCNN (Lai et al., 2015) 47.21 - - - - 96.49 S-LSTM (Zhu et al., 2015) - 81.9 - - - - LSTM (Tai et…”
Dependency Sensitive Convolutional Neural Networks for Modeling Sentences and Documents (2016, methodology): “(Yin and Schütze, 2015) further utilizes multichannel embeddings and unsupervised pretraining to improve classification results.”
MGNC-CNN: A Simple Approach to Exploiting Multiple Word Embeddings for Sentence Classification (2016, methodology): “Our approach enjoys the following advantages compared to the only existing comparable model (Yin and Sch¨utze, 2015): (i) It can leverage diverse, readily available word embeddings with different dimensions, thus providing ﬂexibility.”
Aspect Level Sentiment Classification with Deep Memory Network (2016, background): “…researchers develop denoising autoen-coder (Glorot et al., 2011), convolutional neural network (Kalchbrenner et al., 2014; Kim, 2014; Yin and Sch¨utze, 2015), sequence based recurrent neural models (Sutskever et al., 2014; Kiros et al., 2015; Li et al., 2015b) and tree-structured…”
Deep learning for extracting protein-protein interactions from biomedical literature (2017, unknown): “In NLP research, such models often use separate channels for different word embeddings (Yin and Sch¨utze, 2015; Shi et al., 2016).”
Team Bertha von Suttner at SemEval-2019 Task 4: Hyperpartisan News Detection using ELMo Sentence Representation Convolutional Network (2019, background): “Most state-of-the-art neural network approaches for document classification use a token sequence as network input (Kim, 2014; Yin and Schütze, 2016; Conneau et al., 2016).”
Bridging CNNs, RNNs, and Weighted Finite-State Machines (2018, methodology): “These representations are then often summed, averaged, or max-pooled to produce a document-level representation (Kim, 2014; Yin and Sch¨utze, 2015).”
A Survey of Optimization Methods From a Machine Learning Perspective (2019, background): “Particularly, supervised learning is further divided into the classiﬁcation problem (e.g., sentence classiﬁcation [10], [50], image classiﬁcations [51], [52], [53], etc.) and regression problem; unsupervised learning is divided into clustering and dimension reduction [54], [55], [56], among others.”