Serialized Graph Parsing for Uniform Meaning Representation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing meaning representation parsing methods struggle to uniformly and accurately parse various types of meaning representation graphs, leading to reduced parsing accuracy when targeting specific graphs.
Innovation Solution
A meaning representation parsing system and method that generates a token array representing a graph structure, including nodes and edges, using a serialized graph generating unit and converting unit to ensure uniform and accurate parsing across different types of graphs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a Transition-based Parser is used to accurately parse specific types of meaning representation graphs, then parsing accuracy is improved, but the system cannot uniformly parse various types of graphs and requires redefining actions for each graph type
Solution Approach 1:
The patent applies universality by designing a unified action set that can parse multiple types of meaning representation graphs (AMR, TAC-KR, WebNLG) without requiring separate action definitions for each graph type. The parser uses a single standardized action vocabulary to handle diverse graph structures, achieving both high accuracy and broad adaptability across different meaning representation formats.
2Adaptability or versatility
If a unified action set is used to parse all meaning representation graphs, then adaptability is improved, but parsing accuracy is significantly reduced compared to specialized parsers
Solution Approach 1:
The patent applies parameter changes by modifying the representation parameters of graph structures to enable unified parsing. Specifically, it transforms different meaning representation graphs into a standardized token array format with consistent node and edge representations, allowing a single parser with fixed action set to achieve high accuracy across diverse graph types without sacrificing adaptability.
3Speed
If graph structures are directly parsed without serialization, then parsing speed is improved, but the system cannot handle various types of meaning representation graphs uniformly
Solution Approach 1:
The patent applies preliminary action by performing graph serialization before parsing. The system first converts diverse meaning representation graphs into a standardized serialized token array format, which preserves the original graph structure while enabling uniform processing. This preliminary transformation step allows the subsequent parsing phase to operate efficiently on a consistent input format, achieving both speed and adaptability.
Data Source
AI summary
A meaning representation parsing system 100 includes: an input unit 110 that accepts input data 210 in a text or a graph; and a serialized graph generating unit 120 that generates a token array (serialized graph 220) representing a graph structure corresponding to the input data 210. Furthermore, the token array generated by the serialized graph generating unit 120 includes at least a first token indicating a node in the graph structure corresponding to the input data 210 and a second token indicating an edge representing a relationship between the nodes.