Topic Tree Generation Using Ontology-Based Keyword Parsing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital information classification systems are slow and often compromise performance for accuracy, failing to efficiently generate topic trees for real-time analytics.
Innovation Solution
The system employs a parsing module to extract keywords from digital information, compares them to an ontology, and sorts the extracted hierarchies using weighted vectors and deep-learning techniques to generate a topic tree, effectively improving classification efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If traditional classification systems are used, then accuracy may be maintained, but processing speed deteriorates and real-time analytics become infeasible
Solution Approach 1:
The patent segments the classification task into multiple independent components: (1) extracting nouns from digital information, (2) removing proper nouns, (3) generating similar words, (4) comparing keywords to ontology, (5) extracting hierarchies, and (6) sorting by relevance. This segmentation allows parallel processing of different aspects of classification, significantly improving processing speed while maintaining accuracy through systematic evaluation of each component.
Solution Approach 2:
The system performs preliminary actions by pre-processing digital information to extract and filter nouns before comparison with the ontology. Proper nouns are removed in advance, and similar words are generated beforehand, so that the actual classification comparison operates on refined, pre-processed data, reducing the computational burden during real-time classification.
2Measurement precision
If comprehensive keyword extraction is performed, then classification accuracy improves, but processing time increases
Solution Approach 1:
The patent extracts only the essential elements (nouns) from digital information for classification, deliberately excluding proper nouns that do not contribute to topic classification. This selective extraction reduces the volume of data requiring further processing while retaining the key semantic elements needed for accurate classification.
Solution Approach 2:
The system generates similar words for each extracted noun, creating a multi-functional keyword set that serves multiple purposes: direct matching with ontology, expanding the semantic coverage, and providing alternative matching paths. This universal approach improves accuracy without proportionally increasing processing time by leveraging the generated similar words across multiple comparison scenarios.
3Measurement precision
If deep-learning techniques are applied, then classification accuracy improves, but system complexity increases
Solution Approach 1:
The patent introduces an intermediary ontology layer between the extracted keywords and the final classification result. The ontology serves as a structured mediator that organizes knowledge domains and relationships, allowing the system to leverage deep-learning techniques for generating similar words and determining relevance while maintaining interpretability through the structured ontology framework.
Solution Approach 2:
The system transforms the classification problem from direct keyword-matching into a multi-dimensional evaluation by comparing keywords against the ontology hierarchy and sorting results by relevance scores. This dimensional transformation allows deep-learning techniques to operate in the relevance space rather than directly in the classification space, improving accuracy while managing complexity through the structured ontology dimension.
Data Source
AI summary
The disclosed computer-implemented method for generating a topic tree for digital information may include parsing the digital information and extracting a set of keywords. This method may also include comparing the set of keywords to an ontology and extracting hierarchies from the ontology that match the set of keywords. The extracted ontology entries may then be pruned and sorted. Various other methods, systems, and computer-readable media are also disclosed.