Cause-Effect Mining via Semantic Co-occurrence Frequency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data mining methods face challenges in identifying cause-effect relationships between objects with different scales and units of measurements, and are limited by information gaps and the need for expert knowledge and customized dictionaries in natural language processing.
Innovation Solution
The method uses co-occurrence frequency measurements of semantic terms to determine object similarities, employing frequent item sets and association rule mining, which are dimensionless and tolerant of missing data, allowing for the examination of cause-effect linkages across disparate environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If natural language processing methods are used to infer cause effect relationships, then information gaps can be addressed, but expert knowledge and customized dictionaries are required
Solution Approach 1:
The patent uses freely available, pre-existing text corpora (such as PubMed, DrugBank, and general Wikipedia) as disposable information sources. These existing text resources are mined directly without requiring custom dictionary construction or expert-knowledge-based processing, thereby reducing system complexity while maintaining reliability in identifying cause-effect relationships.
2Adaptability or versatility
If observations with different scales and units of measurements are analyzed, then broader relationships can be identified, but measurement compatibility becomes difficult
Solution Approach 1:
The patent extracts only the semantic meaning and co-occurrence patterns from observations, deliberately removing scale and unit information. By focusing solely on the presence/absence and frequency of semantic terms across different contexts, the system achieves cross-environment adaptability without being constrained by measurement compatibility issues.
Solution Approach 2:
The patent transforms quantitative measurement data into qualitative semantic term frequencies. By changing the parameter representation from measured values with units to dimensionless co-occurrence counts of semantic terms, the system enables comparison across different scales and units while maintaining analytical precision.
3Measurement precision
If frequent observations are used for data mining, then statistical significance is improved, but meaningful but infrequent observations are missed
Solution Approach 1:
The patent implements a dynamic thresholding approach where the significance threshold adapts based on the frequency distribution of semantic terms. Rare but consistently co-occurring terms across different contexts are identified as significant, allowing the system to capture infrequent but meaningful observations while maintaining statistical rigor through context-aware frequency analysis.
Data Source
AI summary
The disclosed embodiments relate to data mining methods for determining economically valuable cause effect relationships between objects and properties associated with objects using co-occurrence frequency measurements of semantic terms characterizing observations of properties, effects or behaviors of objects in different environments and using these measurements as object descriptors in calculations determining object similarities. Specifically, these methods may be used to identify new indications of medicines, identify biomarkers associated with disease, identify biomarkers associated with drug effects, quantify disease diagnosis, identify novel drug targets, identify pharmacologic equivalencies of medicines, identify pharmacologic equivalencies between medicines and traditional medicines, identify pharmacologic equivalencies between medicines and Natural products, identify equivalencies between alternate medical procedures, identify risk benefit profiles of medicine combinations, identify targets for antibodies, identify synergies between medicines, identify Side effects of medicines, identify risks of experimental medicines, identify functions of biological networks.