Fault-Tolerant Information Extraction Using Grammar and External Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current information extraction methods, particularly for named-entity recognition, face challenges in achieving high precision and recall due to limitations in handling noisy and unstructured natural-language inputs, such as typographical errors and morphological variations, which traditional grammar-based and statistical approaches struggle to address effectively.
Innovation Solution
A computer-implemented method utilizing an augmented transition network with external functions that allow for read/write access to memory locations, enabling the evaluation of external functions attached to state transitions, which can handle morphology, syntax, and semantic issues, and provide measures for text similarity, thus improving the extraction process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If rule-based or linguistic grammar-based techniques are used for information extraction, then precision is improved, but recall deteriorates when testing data contains unknown expressions or violations of standard grammatical rules
Solution Approach 1:
The invention segments the information extraction process into two distinct components: a grammar-based analysis component that ensures precision through structural rules, and a statistical analysis component that improves recall by learning from annotated data. These components operate in parallel and their results are combined, allowing the system to maintain high precision while achieving high recall even with noisy input data
Solution Approach 2:
The invention merges rule-based grammar techniques with statistical machine learning approaches into a unified information extraction system. The grammar model provides structural constraints for precision, while the statistical model handles variability and noise for recall, creating a hybrid system that achieves both high precision and high recall simultaneously
2Reliability
If statistical models with Machine Learning are used for information extraction, then recall is improved, but precision deteriorates compared to grammar-based systems
Solution Approach 1:
The system segments the analysis into statistical processing that captures patterns for high recall, and grammar-based processing that enforces structural rules for high precision. Each component addresses the weaknesses of the other, with the statistical model handling unknown expressions and the grammar model ensuring structural accuracy
Solution Approach 2:
The invention implements feedback mechanisms where the results from statistical analysis are validated against grammar rules, and vice versa. This mutual validation process allows the system to maintain high recall from statistical learning while ensuring precision through grammar-based verification of extracted entities
3Measurement precision
If traditional grammar-based approaches are used, then precision is maintained, but the system cannot tolerate morphology, syntax, semantic, or discursive issues in natural language input
Solution Approach 1:
The invention changes the operational parameters of the grammar-based system by integrating it with statistical models that are trained on diverse annotated data including noisy and erroneous inputs. This allows the system to adapt to morphological, syntactic, semantic, and discursive variations while maintaining precision through the grammar component's structural validation
4Reliability
If statistical approaches using large amounts of annotated data are used, then recall is improved for unknown expressions, but the system requires extensive training data and currently yields lower precision than grammar-based approaches
Solution Approach 1:
The invention merges statistical machine learning that leverages large annotated datasets for high recall with grammar-based methods that provide structural precision. The combined system uses the statistical model to identify potential entities from training data while the grammar model validates these detections against linguistic rules, achieving both high recall for unknown expressions and high precision
Data Source
AI summary
A computer-implemented method for automatically analyzing a natural language input for information extraction comprises (i) a step of receiving the natural language input; (ii) a step of providing a grammar model comprising: a local grammar model, a set of external functions, and a finite set of read/write shared memory registers used by a parsing engine and the external functions; (iii) a step of applying the grammar model to the natural language input using the parsing engine, and (iv) a step of extracting information from the natural language input using at least one new output of the grammar model, the new output of the grammar model being built based: on at least one return value of the external functions from evaluating the one or more external functions in step (iii), and one or more input labels and/or output labels.