Text Parsing Using Mutual Information Scores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing natural language parsing techniques are slow due to their reliance on exhaustive search strategies, despite advancements in prioritizing node formation rules and metrics for faster parse tree construction.
Innovation Solution
A method utilizing a statistical trigram mutual information score to evaluate nodes during parsing, based on the correlation between phrase levels and word classes, to guide the construction of parse trees more efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If exhaustive search strategies are used to build all possible full parse trees, then parsing accuracy is improved, but parsing speed deteriorates
Solution Approach 1:
The patent pre-calculates mutual information scores for all possible node combinations before actual parsing begins. These scores are stored in a lookup table, allowing the parser to quickly retrieve and compare pre-evaluated node quality metrics during parsing without performing exhaustive searches, thus maintaining accuracy while improving speed
Solution Approach 2:
The patent changes the evaluation parameter from simple statistical metrics to mutual information scores that measure the information shared between a node and the entire text segment. This parameter change enables more accurate identification of high-quality parse nodes, reducing the number of candidate nodes that need to be explored while maintaining parsing accuracy
2Speed
If prioritizing techniques are used to order rules and nodes, then parsing speed is improved, but parsing accuracy may deteriorate
Solution Approach 1:
The patent uses mutual information scores as a feedback mechanism to guide the prioritization of nodes during parsing. Nodes with higher mutual information scores (indicating they share more information with the text segment and are more likely to be correct) are prioritized for expansion. This feedback-driven prioritization maintains accuracy by focusing on high-quality nodes while improving speed by avoiding low-quality candidates
Data Source
AI summary
A method and parser are provided that generate a score for a node identified during a parse of a text segment. The score is based on a mutual information score that measures the mutual information between a phrase level for the node and a word class of at least one word in the text segment.