Cascaded Machine Learning for Spatial Computing Data Visualization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
IVR call executives face challenges in providing efficient and satisfactory assistance due to a lack of easily accessible reference material, hindering their ability to help callers effectively.
Innovation Solution
A system and method utilizing cascading machine learning models for data visualization on spatial computing devices, which includes language classification, intent recognition, and keyword extraction to automate the selection of relevant data records for display, enhancing interaction efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual searching through extensive databases is performed, then relevant information can be found, but time consumption and resource utilization increase significantly
Solution Approach 1:
The system performs preliminary actions by pre-processing communication transcripts to extract keywords, identify language, and determine intent before the actual information retrieval process. This allows the system to quickly filter and present relevant data records without manual searching, resolving the contradiction between accurate information retrieval and time consumption.
Solution Approach 2:
The patent introduces an intermediary processing layer that includes language classification, intent identification, and keyword extraction models. This intermediary automatically bridges the gap between raw communication data and relevant information records, eliminating the need for manual searching while maintaining high accuracy in information retrieval.
2Measurement precision
If multiple machine learning models are cascaded for comprehensive analysis, then accuracy and personalization improve, but system complexity increases
Solution Approach 1:
The complex processing system is segmented into multiple specialized machine learning models, each handling a specific task: language classification, intent identification, and keyword extraction. This segmentation allows the system to achieve high accuracy through comprehensive analysis while managing complexity by dividing functions into modular, independent components.
Solution Approach 2:
The cascaded machine learning models serve multiple functions within a unified processing pipeline. The same processing system performs language identification, intent analysis, and keyword extraction, making the complex system universally applicable to various communication types while maintaining streamlined operations.
3Productivity
If automated selection of data records is implemented, then productivity and resource allocation improve, but system complexity increases
Solution Approach 1:
The system implements self-service automation where machine learning models automatically select and prioritize data records based on extracted keywords and identified intent, without requiring manual intervention. This self-service capability improves productivity and resource allocation while the modular architecture keeps the automation system manageable through clear separation of concerns.
Data Source
AI summary
The disclosed system utilizes a language classifier to identify the language of a communication transcript. Additionally, the system identifies the intent of the communication transcript using a first natural language processing (NLP) MLM, detects keywords within the communication transcript using a second NLP MLM, and selects target data records to associate with the communication transcript. The selection of target data records includes determining, for each data record from a set of data records, a probability score indicating the likelihood that each data record is associated with the communication transcript. This probability score is determined by a data record prediction MLM. Moreover, the system selects target data records with the highest-ranking probability scores. Finally, the system generates a display report based on the selected target data records and renders the display report on a display of a spatial computing device.