Abstract:Fine-grained sentiment transfer demands to edit an input sentence on a given sentiment intensity while preserving its content, which largely extends traditional binary sentiment transfer. Previous works on sentiment transfer usually attempt to learn latent content representation disentangled from sentiment. However, it is difficult to completely separate these two factors and it is also not necessary. In this paper, we propose a novel model that learns the latent representation without disentanglement and leverages sentiment intensity as input to decoder for fine-grained sentiment control. Moreover, aligned sentences with the same content but different sentiment intensities are usually unavailable. Due to the lack of parallel data, we construct pseudo-parallel sentences (i.e, sentences with similar content but different intensities) to relieve the burden of our model. In specific, motivated by the fact that the sentiment word (e.g., "delicious") has a close relationship with the non-sentiment context word (e.g., "food"), we use dependency parsing to capture the dependency relationship. The pseudo-parallel sentences are produced by replacing the sentiment word with a new one according to the specific context word. Besides, the difference between pseudo-parallel sentences and generated sentences and other constraints are utilized to guide the model precisely revising sentiment. Experiments on the Yelp dataset show that our method substantially improves the degree of content preservation and sentiment accuracy and achieves state-of-the-art performance.

Enhancing Unsupervised Generative Dependency Parser With Contextual

Combining Generative and Discriminative Approaches to Unsupervised Dependency Parsing via Dual Decomposition

Lexicalized Neural Unsupervised Dependency Parsing

Enhancing Unsupervised Semantic Parsing with Distributed Contextual Representations

A Semi-Autoregressive Graph Generative Model for Dependency Graph Parsing

Coupling Retrieval and Meta-Learning for Context-Dependent Semantic Parsing

Second-Order Unsupervised Neural Dependency Parsing

A Survey of Unsupervised Dependency Parsing.

A Generative Parser with a Discriminative Recognition Algorithm

Probabilistic Transformer: A Probabilistic Dependency Model for Contextual Word Representation

Self-attentive Biaffine Dependency Parsing.

Unsupervised Dependency Graph Network

A Neural Probabilistic Structured-Prediction Model for Transition-Based Dependency Parsing.

APGN: Adversarial and Parameter Generation Networks for Multi-Source Cross-Domain Dependency Parsing

Transition-based Dependency Parsing Using Two Heterogeneous Gated Recursive Neural Networks.

How Far Are We from Effective Context Modeling ? an Exploratory Study on Semantic Parsing in Context.

Ambiguity-aware Ensemble Training for Semi-supervised Dependency Parsing.

Semi-Supervised Feature Transformation for Dependency Parsing.

Improve Discourse Dependency Parsing with Contextualized Representations

Fine-Grained Text Sentiment Transfer Via Dependency Parsing

CRF Autoencoder for Unsupervised Dependency Parsing