Abstract:The joint extraction of elements such as entities, relations, events, and their arguments, along with their specific interrelationships, is a critical task in natural language processing. Existing research often employs implicit handling of interactions between tasks through techniques like shared encodings or parameter sharing, lacking explicit modeling of the specific relationships between tasks. This limitation hinders the full utilization of inter-task correlation information and impacts effective collaboration between tasks. To address this, we propose a model for joint extraction of elements and relations based on causal relationship representation enhancement (CRE). This model aims to capture specific relationships between tasks in multiple stages, facilitating finer adjustments and optimization of subtasks and thereby enhancing the overall model performance. Specifically, CRE comprises three key modules: feature adaptation, feature interaction, and feature fusion. The feature adaptation module selects and adjusts features from shared encodings based on the requirements of specific tasks to better adapt to semantic differences between different tasks. The feature interaction module employs causal reasoning to comprehensively capture the causal relationships between tasks while mitigating negative transfer brought about by interference from semantic information. The feature fusion module further integrates features to obtain optimized task-specific representations. Ultimately, CRE exhibits a significant improvement in average performance across multiple information extraction tasks.

To Extract Concept Instances and Concept Attributes Based on Coordinate Structure

Exploiting Multi-Category Characteristics and Unified Framework to Extract Web Content

Learning to Mine Chinese Coordinate Terms Using the Web

A Head-Driven Approach For Automatic Identification Of Chinese Coordinate Structures

Automatic Extraction Of Commodity Attributes On Webpages Based On Hierarchical Structure

Exploring and Verbalizing Academic Ideas by Concept Co-occurrence

An Integrated Approach for Concept Learning and Relation Extraction

Induction of Semantic Classes Based on Coordinate Patterns

Weakly-Supervised Extraction of Ontology Concept Instances and Concept Attributes from the Web

Part Ontology Mapping in Cooperative Enterprises

Large-scale Automatic Extraction of Chinese Compound Lexical Cohesion Pairs

Resolving Coordinate Structures for Chinese Constituent Parsing.

Causal-relationship representation enhanced joint extraction model for elements and relationships

A Model for Extraction of the Concept of Chinese General Ontology Based on Pattern Matching

Unsupervised Extraction of Representative Concepts from Scientific Literature

An efficient and effective approach for multi-fact extraction from text corpus

Bootstrapping Information Extraction Via Conceptualization

Extraction and reuse of pattern configuration based on ontology and shape grammar

Extracting hyponymy relation between Chinese terms based on term types'commonality and sequential patterns

A Two Dimensional Feature Engineering Method for Relation Extraction

FCDS: Fusing Constituency and Dependency Syntax into Document-Level Relation Extraction