Core 1: Knowledge Extraction
To extract knowledge from multiple resources including structured and unstructured, represent it in the form of Ripple Down Rules (RDR) which is easy to use and evolvable with minimum efforts.
- Extract hidden knowledge from the structured data using cost sensitive feature selection, evolutionary computing and machine learning algorithms.
- Process unstructured contents using state of the art NLP techniques and extract actionable knowledge.
- Develop a system to process images and provide preliminary visual knowledge (Image type, organ kind, defect status).
- Design and develop methodology for Context-aware dialogue management.
- All the knowledge extracted are initially stored in the plain rules format.
- Cases are generated to adapt the extracted plain rules in Ripple Down Rules (RDR) format
- Accurate inference over an incremental knowledge base with competent accuracy by exploiting just-in-time verification and validation
Uniqueness & Contribution
- Unified Score based valuable features selection
- Dynamic algorithm selection based on input data
- Automatic knowledge evolution based on new concept identification
- Use only pixels in the allocated Region of Interest (ROI) for next stage of defect detection
- Context-Aware Dialogue Manager (CADM) model for enrichment and completeness using effective conversational service
- Missing values identification in user input while performing inferencing on the RDR knowledge.