Structured Prediction Cascade for Dependency Parsing Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Dependency parsing algorithms face exponential computational complexity as the number of words in a text increases, making them computationally expensive and inefficient, especially for higher-order parsing.
Innovation Solution
A computer-implemented method that employs a structured prediction cascade using pruning algorithms, such as augmented vine parsing and multiple-order parsing models, to reduce the index set of possible head-modifier dependencies, focusing on inner and outer arcs based on distance thresholds and likelihood scores, to efficiently determine the most-likely parse.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a parsing algorithm analyzes and scores each index individually to find the most-likely parse, then the accuracy of the parsing result is improved, but the computational time and complexity increase exponentially
Solution Approach 1:
The patent segments the index set into inner arcs and outer arcs based on distance thresholds. This segmentation allows different processing strategies to be applied to different parts of the search space, reducing the overall computational burden while maintaining parsing accuracy through targeted analysis of the most promising dependencies.
Solution Approach 2:
The patent applies preliminary pruning actions to eliminate unlikely arcs before the main parsing algorithm processes the full index set. By pre-filtering the search space using distance-based criteria and training data statistics, the system reduces the number of indices that require expensive individual scoring, thereby decreasing computational time while preserving accuracy for the remaining candidates.
2Reliability
If the parsing algorithm considers all possible head-modifier dependencies in the index set, then the completeness of the parse is improved, but the number of computations becomes prohibitively expensive
Solution Approach 1:
The patent extracts and removes unlikely arcs from the index set through pruning operations based on distance thresholds and training data statistics. By taking out these low-probability candidates before the main parsing process, the system maintains parsing completeness for the remaining relevant arcs while dramatically improving computational efficiency by eliminating unnecessary computations.
Solution Approach 2:
The patent changes the parameter of the index set by applying transformations that reduce its size through pruning. By modifying the index set to exclude arcs that exceed distance thresholds or fail statistical criteria, the system maintains the essential parsing information while reducing the computational burden from considering all possible n² indices.
3Measurement precision
If higher-order dependency parsing is performed to improve parse accuracy, then the quality of the parsing result is improved, but the number of computations increases even more
Solution Approach 1:
The patent segments the computational process into distinct phases: pruning based on distance thresholds, pruning based on training data statistics, and final parsing. This segmentation allows higher-order dependency parsing to be applied selectively to the pruned, smaller index set rather than the full n² index set, thereby maintaining parse accuracy while reducing overall computational complexity.
Solution Approach 2:
The patent performs preliminary pruning actions using distance-based criteria and statistical filters before applying the computationally intensive higher-order parsing algorithms. This preliminary action reduces the search space to only the most promising arcs, allowing higher-order parsing to achieve improved accuracy without the prohibitive computational cost of applying it to the complete index set.
Data Source
AI summary
A dependency parsing method can include determining an index set of possible head-modifier dependencies for a sentence. The index set can include inner arcs and outer arcs, inners arcs representing possible dependency between words in the sentence separated by a distance less than or equal to a threshold and outer arcs representing possible dependency between words in the sentence separated by a distance greater than the threshold. The index set can be pruned to include: (i) each specific inner arc when a likelihood that the specific inner arc is appropriate is greater than a first threshold, and (ii) the outer arcs when a likelihood that there exists any possible outer arc that is appropriate is greater than the first threshold. The method can include further pruning the pruned index set based on a second parsing algorithm, and determining a most-likely parse for the sentence from the pruned index set.