Real-Time Event Interpretation With Symbolic-AI Mediation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing software systems struggle with processing complex data such as natural language due to limitations in both Symbolic-Algorithmic Systems, which lack flexibility for complex data, and Statistical Machine-Learning Systems, which are costly, unexplainable, and lack precision, making it difficult to handle Heterogeneous and Unreasonably Broad (HUB) applications that require broad, heterogeneous data sets.
Innovation Solution
A computer-implemented method using a deep learning model to detect and interpret real-time events in a structured, machine-readable representation of data, enabling the creation of new semantic nodes, processing and expanding this representation to include new knowledge, and performing tasks like reasoning, translation, and question answering within a unified machine-readable language framework.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Symbolic-Algorithmic Systems are used to process complex data, then precision is maintained at 100%, but the ability to handle complex data such as natural language, speech, and images deteriorates
Solution Approach 1:
The patent segments the processing of complex data into two distinct stages: first, a statistical machine-learning system processes unstructured complex data (natural language, speech, images) to extract meaningful information; second, this extracted information is structured and processed by symbolic-algorithmic systems for precise reasoning and decision-making. This segmentation allows each system type to operate in its optimal domain.
Solution Approach 2:
The patent introduces an intermediary layer that bridges statistical machine-learning systems and symbolic-algorithmic systems. This intermediary translates the probabilistic outputs of statistical systems into structured representations that symbolic systems can process, enabling the combination of both approaches while maintaining precision in the final output.
2Adaptability or versatility
If Statistical Machine-Learning Systems are used to process complex data, then the ability to handle heterogeneous data improves, but precision deteriorates from 100% to below 100%
Solution Approach 1:
The patent segments the processing pipeline so that statistical machine-learning systems handle only the initial processing of complex data to extract features and patterns, while the final decision-making and precision-critical operations are performed by symbolic-algorithmic systems that operate on structured representations of this extracted information.
Solution Approach 2:
The patent implements feedback mechanisms where the outputs of statistical machine-learning systems are continuously evaluated and refined before being passed to symbolic systems. This feedback loop allows the system to learn from errors and improve precision while maintaining the ability to process complex heterogeneous data.
3Adaptability or versatility
If Deep Learning models with hundreds of billions of parameters are used, then the ability to process complex data improves, but cost deteriorates to 10's or 100's of millions of US$
Solution Approach 1:
The patent extracts only the essential features and patterns from complex data using relatively small statistical machine-learning models, rather than using enormous deep learning models to process all raw data. This extraction approach maintains the ability to handle complex data while dramatically reducing computational costs and resource requirements.
Solution Approach 2:
The patent employs lightweight, computationally efficient models for data processing that can be quickly trained and deployed, replacing the need for expensive, large-scale deep learning models. These simpler models achieve sufficient performance for the specific tasks at hand while costing a fraction of the resources.
Data Source
AI summary
There is provided a computer implemented method in which a deep learning model detects and interprets real time events from an input data stream, in which the detected and interpreted events are output in a structured, machine-readable representation of data that conforms to a machine-readable language.