Learning Distributed Representations of Symbolic Structure Using Binding and Unbinding Operations
Explore this paper's citation graph
Summary
This work proposes the TPRU, a recurrent unit that, at each time step, explicitly executes structural-role binding and unbinding operations to incorporate structural information into learning.
- Type
- preprint
- Published
- 2018-10-29
- Cited by
- 4
- References
- 66
- Access
- Open access
- OpenAlex
- https://openalex.org/W2898996181
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:53113804
Keywords
Task (project management), Computer science, Natural language processing, Artificial intelligence, Inference
References
- On the difficulty of training recurrent neural networks
- Neural Machine Translation of Rare Words with Subword Units
- A large annotated corpus for learning natural language inference
- Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
- Recursive Distributed Representations
- Unsupervised Construction of Large Paraphrase Corpora: Exploiting Massively Parallel News Sources
- Annotating Expressions of Opinions and Emotions in Language
- Symbolic functions from neural computation
- Long Short-Term Memory
- Learning Question Classifiers
- Learning Classification with Unlabeled Data
- A Sentimental Education: Sentiment Analysis Using Subjectivity Summarization Based on Minimum Cuts
- SemEval-2014 Task 10: Multilingual Semantic Textual Similarity
- SemEval-2015 Task 2: Semantic Textual Similarity, English, Spanish and Pilot on Interpretability
- Backpropagation Through Time: What It Does and How to Do It
- *SEM 2013 shared task: Semantic Textual Similarity
- Mining and summarizing customer reviews
- Seeing Stars: Exploiting Class Relationships for Sentiment Categorization with Respect to Rating Scales
- Representation Learning: A Review and New Perspectives
- A SICK cure for the evaluation of compositional distributional semantic models
Cited by
- RNNs Implicitly Implement Tensor Product Representations
- Enhancing the Transformer with Explicit Relational Encoding for Math Problem Solving
- Extracting Relational Explanations From Deep Neural Networks: A Survey From a Neural-Symbolic Perspective
- Tensor Product Representation Probes Reveal Shared Structure Across Linear Directions
Related papers
- Sequence learning with recurrent networks: analysis of internal representations
- Linguistic Knowledge as Memory for Recurrent Neural Networks
- Clustering and compositionality of task representations in a neural network trained to perform many cognitive tasks
- Distributed representations and nested compositional structure
- Emergent latent symbol systems in recurrent neural networks
- Hierarchical Reasoning with Distributed Vector Representations
- DRAGNN: A Transition-based Framework for Dynamically Connected Neural Networks
- Recurrent Neural Network for Syntax Learning with Flexible Representations
- Learning Associative Inference Using Fast Weight Memory