Automatic Rule Generation for Information Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for generating information extraction rules are complex, time-consuming, and require specialized skills, especially for non-common entities and events, and are limited by the need for abundant data occurrences in the corpus, making it difficult for domain experts with little expertise to develop accurate and efficient extraction rules.
Innovation Solution
A computer-implemented system with modules for rule generation, relaxation, testing, and candidate suggestion that iteratively refines extraction rules using untagged data, allowing users to generate rules without manual writing and reducing the effort required for sample data collection, enabling the development of accurate rules for non-common entities and events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If bootstrapping-based systems are used to generate extraction rules automatically, then the effort required for rule development is reduced, but the systems fail to work for less common entities and events due to lack of abundant data occurrences
Solution Approach 1:
The patent introduces an intermediary component that connects bootstrapping-based automatic rule generation with less common entities and events. This intermediary enables the system to handle rare entities by acting as a bridge between the automated generation process and the specific requirements of less common extraction tasks, thereby maintaining both ease of rule development and adaptability to diverse entity types.
2Measurement precision
If manual rule writing is performed by domain experts, then accurate extraction rules can be developed, but the process is complex and time-consuming
Solution Approach 1:
The patent applies preliminary action by pre-processing and preparing data structures, patterns, and templates before the actual rule generation process. This preliminary preparation enables the automatic generation system to produce accurate extraction rules more quickly, reducing the time required while maintaining the precision that would otherwise require extensive manual rule writing by domain experts.
3Ease of operation
If Machine Learning systems are used to generate extraction rules, then specialized skills are not required, but large amounts of tagged data are needed which requires significant effort to collect
Solution Approach 1:
The patent extracts and separates the tagging process from the rule generation process. Instead of requiring large amounts of pre-tagged data as input, the system extracts relevant information directly from untagged text during rule generation, eliminating the need for extensive manual tagging while maintaining ease of operation for non-expert users.
4Measurement precision
If extraction rules are optimized for accuracy, then the rules may become complex and require too many system resources to run
Solution Approach 1:
The patent employs parameter changes by dynamically adjusting rule complexity parameters based on the specific extraction task and available system resources. The system can modify parameters such as rule depth, pattern complexity, and processing thresholds to achieve optimal precision while controlling the computational resources required, thereby balancing accuracy with device complexity.
Data Source
AI summary
An automatic rule generation system generates rules for fact extraction. A rule generation module receives a sample and generates a rule from the sample. A rule relaxation module generates a relaxed rule from the rule. A rule testing module generates a reverse index from a corpus, applies the relaxed rule to the reverse index, and generates text segments. An information extraction module generates modified text segments from the relaxed rule and the text segments. A candidate suggestion module performs a candidate generation process: if the candidate generation process generates no candidates, the candidate suggestion module signals the rule relaxation module to generate a further relaxed rule to use as the relaxed rule. A user evaluates a candidate and provides the candidate as an additional sample for the automatic rule generation system to generate another rule to use as the rule. As a result of performing these actions iteratively, the rule is eventually generated and relaxed to result in an appropriate rule to use for fact extraction.