【論文閱讀筆記】Aspect-based sentiment analysis with alternating coattention networks
Aspect-based sentiment analysis with alternating coattention
networks
目錄
- introduction
- method
- experiments
- conclusion and future work
introduction
ABSA任務包括兩個子任務:方面檢測和情感分析。
Coattention-LSTM
Coattention-MemNet
Location-enhanced coattention
method
an overview of coattention networks
Coattention-LSTM:
context and target encoder:
coattention encoder:
coattention encoder:
Coattention-Memnet:
context and target encoder:
coattention encoder:
sentiment analysis:
與共Coattention- LSTM相似，使用softmax層作為分類器，交叉熵函式作為損失函式。
Location-enhance coattention:
Great food but the service was dreadful!
experiments
資料集：SemEval 2014 和 Twitter
評價指標：
結果:
conclusion and future work
由於我們的方法使用LSTM或單詞嵌入層來表示單詞，注意只是一個加權和函式，因此很難學習到否定修飾詞與隱含的情感短語之間的複雜關係。因此，我們需要改進LSTM或attention，使網路學習上下文詞之間的複雜關係，或將注意層的結果以非線性的方式對映到分類空間。
在未來的工作中，作者計劃設計一個上下文學習函式來捕捉區域性上下文單詞之間的關係，該關係的複雜度在單詞嵌入和LSTM之間。我們還將考慮在網路中加入外部知識來解決否定修飾詞建模和識別未知情緒詞和短語的問題。
相關文章
- [論文閱讀筆記] Adversarial Learning on Heterogeneous Information Networks筆記ORM
- 論文閱讀筆記：Fully Convolutional Networks for Semantic Segmentation筆記Segmentation
- MapReduce 論文閱讀筆記筆記
- SSD論文閱讀筆記筆記
- AutoEmbedding論文閱讀筆記筆記
- Image Super-Resolution Using DeepConvolutional Networks論文閱讀筆記筆記
- 論文閱讀——Deformable Convolutional NetworksORM
- [論文閱讀筆記] metapath2vec: Scalable Representation Learning for Heterogeneous Networks筆記
- CornerNet-Lite論文閱讀筆記筆記
- Delphi 論文閱讀 Delphi: A Cryptographic Inference Service for Neural Networks
- 論文閱讀 TEMPORAL GRAPH NETWORKS FOR DEEP LEARNING ON DYNAMIC GRAPHS
- 閱讀論文：《Compositional Attention Networks for Machine Reasoning》Mac
- [論文閱讀筆記] Structural Deep Network Embedding筆記Struct
- 論文閱讀 A Data-Driven Graph Generative Model for Temporal Interaction Networks
- 論文閱讀：《Deep Compositional Question Answering with Neural Module Networks》
- [論文閱讀] Patient subtyping via time-aware LSTM networks
- 【論文筆記】Shortest Paths and Centrality in Uncertain Networks筆記AI
- Reading Face, Read Health論文閱讀筆記筆記
- ACL2020論文閱讀筆記：BART筆記
- 兩篇知識表示方面的論文閱讀筆記筆記
- 論文閱讀：SiameseFC
- 【論文閱讀筆記】An Improved Neural Baseline for Temporal Relation Extraction筆記
- 論文閱讀：《Multimodal Graph Networks for Compositional Generalization in Visual Question Answering》
- DMCP: Differentiable Markov Channel Pruning for Neural Networks 閱讀筆記筆記
- Raft論文讀書筆記Raft筆記
- [論文閱讀筆記] Community aware random walk for network embedding筆記Unityrandom
- 論文閱讀筆記：A Two-Step Approach for Event Factuality Identification筆記APPIDE
- 如何閱讀科研論文
- 論文解讀（GAN）《Generative Adversarial Networks》
- 並行多工學習論文閱讀（五）：論文閱讀總結並行
- 《End-to-End Adversarial Memory Network for Cross-domain Sentiment Classification》閱讀筆記ROSAI筆記
- [論文閱讀筆記] Adversarial Mutual Information Learning for Network Embedding筆記ORM
- 論文閱讀筆記：LINE: Large-scale Information Network Embedding筆記ORM
- 論文閱讀筆記（五）：Hire-MLP Vision MLP via Hierarchical Rearrangement筆記
- 論文閱讀筆記（四）：AS-MLP AN AXIAL SHIFTED MLP ARCHITECTUREFOR VISION筆記
- [論文解讀]A Quantitative Analysis Framework for Recurrent Neural NetworkFramework
- 論文解讀（GIN）《How Powerful are Graph Neural Networks》
- Q-REG論文閱讀