Dependency Tree Pruning for Accurate Subevent Relation Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional natural language processing models for subevent relation extraction fail to identify important context words and filter out irrelevant words due to their reliance on the sequential format of documents, leading to reduced prediction accuracy.
Innovation Solution
The proposed NLP apparatus generates a dependency tree, identifies non-dependency path important words, and performs optimal transport to prune the tree, retaining only relevant context words for improved subevent relation prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional NLP models rely on sequential format of documents, then the model structure is simple, but prediction accuracy is reduced
Solution Approach 1:
The patent segments the document processing into two distinct components: sequential processing for basic understanding and tree-structured processing for contextual relationships. The dependency tree structure divides words into hierarchical levels (root, intermediate, leaf nodes) allowing the model to process contextual information in an organized manner while maintaining structural clarity.
Solution Approach 2:
The patent transitions from one-dimensional sequential processing to two-dimensional tree-structured processing. By organizing words into a dependency tree with hierarchical levels and positional relationships, the model gains an additional dimensional perspective for capturing contextual information, enabling better identification of important context words beyond simple sequence order.
2Productivity
If all context words are retained in the dependency tree, then no information is lost, but processing efficiency is reduced
Solution Approach 1:
The patent extracts only the essential context words from the complete dependency tree based on specific criteria (position relative to event mentions, dependency relationships, and importance scoring). This extraction process removes irrelevant words while preserving the critical contextual information needed for accurate subevent relation extraction, achieving both efficiency and information retention.
Solution Approach 2:
The patent applies different processing qualities to different parts of the dependency tree. Words closer to event mentions receive higher importance weights and are more likely to be retained, while distant words are filtered more aggressively. This local quality differentiation ensures that processing resources are concentrated on the most informative regions of the text.
Data Source
AI summary
A method, apparatus, and non-transitory computer readable medium for natural language processing are described. Embodiments of the present disclosure include obtaining a document comprising a first event mention and a second event mention. Some embodiments generate a dependency tree based on the document. The dependency tree is pruned by removing an irrelevant word to obtain a pruned dependency tree. Subevent relation information is generated for the first event mention and the second event mention based on the pruned dependency tree.