Speaker-Oriented Latent Structures for Dialogue-Based Relation Extraction (2021-09-11T00:00:00.000000Z)

TL;DR

This work introduces SOLS, a novel model which can explicitly induce speaker-oriented latent structures for better DiaRE, and learns latent structures to capture the relationships among tokens beyond the utterance boundaries, alleviating the entangled logic issue.

Abstract

Dialogue-based relation extraction (DiaRE) aims to detect the structural information from unstructured utterances in dialogues. Existing relation extraction models may be unsatisfactory under such a conversational setting, due to the entangled logic and information sparsity issues in utterances involving multiple speakers. To this end, we introduce SOLS, a novel model which can explicitly induce speaker-oriented latent structures for better DiaRE. Specifically, we learn latent structures to capture the relationships among tokens beyond the utterance boundaries, alleviating the entangled logic issue. During the learning process, our speaker-specific regularization method progressively highlights speaker-related key clues and erases the irrelevant ones, alleviating the information sparsity issue. Experiments on three public datasets demonstrate the effectiveness of our proposed approach.

Authors

Wei Lu

8 papers

Guoshun Nan

3 papers

Sicong Leng

2 papers

TL;DR

Abstract

Authors

References77 items

Interventional Video Relation Detection

Uncovering Main Causalities for Long-tailed Information Extraction

PRGC: Potential Relation and Global Correspondence Based Joint Relational Triple Extraction

Document-level Relation Extraction as Semantic Segmentation

AliCG: Fine-grained and Evolvable Conceptual Graph Construction for Semantic Search at Alibaba

Interventional Video Grounding with Dual Contrastive Learning

Semantic Representation for Dialogue Modeling

TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Finance

Video Corpus Moment Retrieval with Contrastive Learning

SUTD-TrafficQA: A Question Answering Benchmark and an Efficient Network for Video Reasoning over Traffic Events

Context-aware Biaffine Localizing Network for Temporal Sentence Grounding

Natural Language Video Localization: A Revisit in Span-Based Question Answering Framework

Document-Level Relation Extraction with Reconstruction

GDPNet: Refining Latent Multi-View Graph for Relation Extraction

Scan2Cap: Context-aware Dense Captioning in RGB-D Scans

Inducing Target-specific Latent Structures for Aspect Sentiment Classification

Document-Level Relation Extraction with Adaptive Thresholding and Localized Context Pooling

Interpreting Graph Neural Networks for NLP With Differentiable Edge Masking

OpenUE: An Open Toolkit of Universal Extraction from Text

Global-to-Local Neural Networks for Document-Level Relation Extraction

Contextualized Perturbation for Textual Adversarial Attack

Contrastive Triple Extraction with Generative Transformer

Dialogue Relation Extraction with Document-Level Heterogeneous Graph Attention Networks

Few-shot Relation Extraction via Bayesian Meta-learning on Relation Graphs

Recurrent Attention Network with Reinforced Generator for Visual Dialog

MIE: A Medical Information Extractor towards Medical Dialogues

Retrieve, Program, Repeat: Complex Knowledge Base Question Answering via Alternate Meta-learning

Reasoning with Latent Structure Refinement for Document-Level Relation Extraction

SciREX: A Challenge Dataset for Document-Level Information Extraction

How Do Decisions Emerge across Layers in Neural Models? Interpretation with Differentiable Masking

Dialogue-Based Relation Extraction

Relation Extraction Exploiting Full Dependency Forests

ScanRefer: 3D Object Localization in RGB-D Scans using Natural Language

Relation Adversarial Network for Low Resource Knowledge Graph Completion

Connecting the Dots: Document-level Neural Relation Extraction with Edge-oriented Graphs

RoBERTa: A Robustly Optimized BERT Pretraining Approach

SpanBERT: Improving Pre-training by Representing and Predicting Spans

Latent Structure Models for Natural Language Processing

Learning Latent Trees with Stochastic Perturbations and Differentiable Dynamic Programming

Attention Guided Graph Convolutional Networks for Relation Extraction

DocRED: A Large-Scale Document-Level Relation Extraction Dataset

Unsupervised Neural Single-Document Summarization of Reviews via Learning Latent Discourse Structure and its Ranking

Inter-sentence Relation Extraction with Document-level Graph Convolutional Neural Network

Enriching Pre-trained Language Model with Entity Information for Relation Classification

Document-Level N-ary Relation Extraction with Multiscale Representation Learning

Learning Dual Retrieval Module for Semi-supervised Relation Extraction

Cooperative Learning of Disjoint Syntax and Semantics

Knowledge Graph Embedding Based Question Answering

Graph Convolution over Pruned Dependency Trees Improves Relation Extraction

N-ary Relation Extraction using Graph-State LSTM

Differentiable Perturb-and-Parse: Semi-Supervised Parsing with a Structured Variational Autoencoder

Graph Convolutional Networks With Argument-Aware Pooling for Event Detection

Simultaneously Self-Attending to All Mentions for Full-Abstract Biological Relation Extraction

Learning Sparse Neural Networks through L0 Regularization

Learning Structured Text Representations

Cross-Sentence N-ary Relation Extraction with Graph LSTMs

Structured Attention Networks

The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables

Question Answering on Freebase via Relation Extraction and Textual Evidence

A Poisson process model for Monte Carlo

From Softmax to Sparsemax: A Sparse Model of Attention and Multi-Label Classification

Classifying Relations via Long Short Term Memory Networks along Shortest Dependency Paths

Semantic Relation Classification via Convolutional Neural Networks with Simple Negative Sampling

GloVe: Global Vectors for Word Representation

Relation Classification via Convolutional Deep Neural Network

Structured Prediction Models via the Matrix-Tree Theorem

2005 Special Issue: Framewise phoneme classification with bidirectional LSTM and other neural network architectures

Bidirectional recurrent neural networks

A simplest systematics for the organization of turn-taking for conversation

BWorld Robot Control Software

NERO: A Neural Rule Grounding Framework for Label-Efficient Relation Extraction

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

Semisupervised classification with graph convolutional networks

Face recognition: a convolutional neural-network approach

S3: I'll take that dad. S1: It seems your daughter and Richard are

The latent structures are then fed to SOLS encoder, which is a graph convolution network (GCN) (Kipf and Welling, 2017) for information