Event Type and Relationship Extraction Based on Dependent Syntactic Semantic Augmented Graph Networks

Abstract

In natural language processing, tasks like trigger word extraction, event type recognition, and event relation extraction are essential. These tasks facilitate the extraction of salient information through detailed textual analysis, enhancing the semantic understanding of texts. The extraction of trigger words and event types often faces challenges due to polysemous words and complex sentence structures, which can impair semantic representation. To address this, this paper introduces a Dependency Syntactic Analysis model and proposes the Event Type Extraction Model based on Gravitational Network with Enhanced Dependency Semantics (GNEDS). This model clarifies complex relationships and structures among words with in sentences, significantly improving contextual information comprehension. This enables more accurate identification of trigger words and their contextual links, thereby enhancing the text’s semantic representation. Furthermore, traditional research in event relation recognition has primarily focused on intrasentential relations, but real-world texts often display multisentence event relations that involve complex contextual and implicit reasoning. To overcome the limitations of existing models in cross-sentence event relation extraction, this study introduces a Graph Convolutional Neural Network (GCN) and an innovative concept of document nodes. A new model, Document Event Relationship Extraction based on Graph Convolutional Network with Enhanced Dependency Semantics (GCNEDS) is proposed. It captures long distance dependencies between sentences within a document with greater accuracy, marking a significant advancement in the field of event type and relationship extraction based on dependent syntactic-semantic augmented graph networks.

Keywords

• Event type extraction
• event relation extraction
• dependent syntax
• GCN
• gravitational network
• BERT