Pruned Dependency Tree for Document-Level Event Argument Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional event argument extraction systems are not scalable for document-level extraction, as they struggle to identify relationships between event trigger words and argument candidate words located far apart in a document, and they fail to effectively prune dependency trees to exclude irrelevant words.
Innovation Solution
The development of an event argument extraction apparatus that uses machine learning techniques, specifically a pruned dependency tree and graph convolutional networks, to predict relationships between event trigger words and argument candidate words, incorporating optimal transport methods for joint optimization of syntactic and semantic distances to model the relationship.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If sentence-level event argument extraction is used, then the relationship between event trigger word and argument candidate word in the same sentence can be determined, but the system is not scalable to perform document-level event argument extraction where argument candidate word can be located far from event trigger word
Solution Approach 1:
The patent segments the dependency tree into useful and irrelevant parts by identifying and removing words that are not part of the dependency path between event trigger and argument candidate. This segmentation enables the system to handle document-level extraction by focusing only on relevant segments while maintaining scalability.
Solution Approach 2:
The patent extracts and removes irrelevant words from the dependency tree, retaining only the essential words that form the dependency path. This extraction process simplifies the input for document-level extraction while preserving the critical relationships between event trigger and argument candidate words.
2Measurement precision
If dependency tree is pruned to exclude irrelevant words, then prediction accuracy is improved, but the complexity of processing the dependency tree increases
Solution Approach 1:
The patent performs preliminary action by pre-identifying and removing irrelevant words from the dependency tree before performing the main extraction task. This preliminary pruning reduces the complexity of subsequent processing while maintaining high prediction accuracy by ensuring only relevant words remain in the tree.
Solution Approach 2:
The patent extracts and removes words that are not part of the dependency path between event trigger and argument candidate. This extraction simplifies the dependency tree structure, making it easier to process while improving prediction accuracy by eliminating noisy or irrelevant information.
3Productivity
If words are removed from dependency tree based on semantic independence, then the number of words to process is reduced, but the risk of removing important words increases
Solution Approach 1:
The patent employs feedback mechanisms where the system continuously evaluates the impact of removing words on the dependency path. By monitoring whether removed words are indeed irrelevant or important, the system adjusts its pruning strategy to maintain reliability while improving processing efficiency.
Solution Approach 2:
The patent replaces simple mechanical word removal with a sophisticated semantic analysis system that evaluates the semantic independence of words. This substitution allows the system to make more informed decisions about which words to remove, reducing the risk of losing important information while maintaining processing efficiency.
Data Source
AI summary
Systems and methods for natural language processing are described. One or more embodiments of the present disclosure generate a word representation vector for each word of a text comprising an event trigger word and an argument candidate word; generate a dependency tree based on the text and the word representation vector; determine that at least one word of the text is independent of a relationship between the event trigger word and the argument candidate word; remove the at least one word from the dependency tree based on the determination to obtain a pruned dependency tree; generate a modified representation vector for each word of the pruned dependency tree using a graph convolutional network (GCN); and identify the relationship between the event trigger word and the argument candidate word based on the modified representation vector for each word of the pruned dependency tree.