Learning Apparatus for Automated Knowledge Base Construction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for acquiring comprehensive knowledge from large numbers of documents are inefficient due to the high cost and resource-intensive process of generating a dictionary for word relationship extraction, limiting the ability to automatically extract relevant information.
Innovation Solution
A learning apparatus and method that uses distant supervision with a small number of rules to label events, generating learning data by extracting event candidates, calculating feature values, and automatically extracting related words to create a knowledge base, enabling efficient generation of learning data for machine learning applications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a known dictionary is used to properly estimate relationships between words, then extraction precision is improved, but device complexity and cost increase due to the work involved in generating the dictionary
Solution Approach 1:
The patent creates simplified copies of dictionary entries by extracting only the necessary relationship patterns from existing dictionaries. Instead of using complete dictionary definitions, it generates compact relationship rules that capture the essential word relationships, reducing the complexity burden while preserving extraction precision.
Solution Approach 2:
The patent performs preliminary actions by automatically generating relationship dictionaries using computational methods before the actual extraction process. This preliminary dictionary generation uses algorithms to infer relationships from corpora, reducing the manual work complexity while providing sufficient precision for accurate word relationship extraction.
2Measurement precision
If a dictionary including knowledge of sufficient quantity is generated, then extraction precision is improved, but loss of time increases due to the extensive work required
Solution Approach 1:
The patent applies partial action by generating only the specific relationship patterns needed for the extraction task, rather than creating a complete comprehensive dictionary. It extracts relationship rules for only the relevant word pairs and concepts required, reducing the time investment while maintaining sufficient precision for the specific extraction objectives.
Solution Approach 2:
The patent changes parameters by adjusting the depth and scope of dictionary generation based on the specific extraction needs. It dynamically controls the quantity and detail level of relationship knowledge generated, optimizing the balance between precision requirements and time constraints for different extraction scenarios.
3Measurement precision
If manual search is performed for comprehensive knowledge acquisition, then extraction precision is improved, but productivity decreases due to limited human resources
Solution Approach 1:
The patent replaces the mechanical manual search process with automated computational systems. It uses machine learning models and algorithmic approaches to perform knowledge acquisition and relationship extraction, maintaining precision through supervised learning while dramatically improving productivity by eliminating human resource constraints.
Solution Approach 2:
The patent enables self-service by creating systems that automatically acquire and organize knowledge without continuous human intervention. The extracted relationships and generated dictionaries can be reused for multiple extraction tasks, allowing the system to serve itself by leveraging previously acquired knowledge for future extraction operations.
Data Source
AI summary
According to one embodiment, a learning apparatus includes a first rule generator, a feature value calculator, a related word extractor, a second rule generator, and a learning unit. The first rule generator generates a first rule to label the event candidate, the first rule including a keyword of the event candidate. The feature value calculator calculates feature values of other words included in the text other than the event candidate. The related word extractor extracts a related word relating to the keyword from the other words using the feature values. The second rule generator generates a second rule to label the event candidate, the second rule being different from the first rule and including the related word. The learning unit generates learning data associating the keyword, the related word, and labeled event candidate with each other.