Cross-Platform Data Insight Assistant Using Large Language Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems are unable to analyze data across different platforms due to the large volume of data and varying data formats, preventing effective insights and automation.
Innovation Solution
An assistant using large language models is trained on user interactions and process graphs to provide insights across automation, process mining, and task mining platforms, integrating AI/ML models for data analysis and automation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional systems are used to store data in separate databases with different formats, then data storage is manageable, but data analysis across platforms becomes impossible
Solution Approach 1:
The patent introduces a language model as an intermediary component that sits between the separate platform databases and the analysis requirements. This language model translates and integrates data from multiple platforms (automation, process mining, task mining) into a unified analytical framework, enabling cross-platform analysis without requiring direct integration of the underlying databases.
Solution Approach 2:
The language model serves multiple functions: it processes natural language queries, retrieves relevant data from different platforms, performs analysis, and generates insights. This multi-functional approach allows a single system component to handle diverse data analysis tasks across platforms that would otherwise require separate specialized systems.
2Loss of information
If large language models are trained on user interactions and process graphs, then comprehensive insights can be provided, but training data requirements and processing time increase
Solution Approach 1:
The patent applies preliminary action by pre-training the language model on extensive user interaction data and process graphs before deployment. This offline training phase prepares the model in advance to handle various query types and data patterns, so that during actual operation, insights can be generated quickly without requiring additional training time.
Solution Approach 2:
The training process is segmented into distinct phases: collecting user interactions, generating process graphs from event logs, training the language model on these structured representations, and then deploying. This segmentation allows the complex training task to be broken into manageable steps that can be performed systematically.
Data Source
AI summary
Systems and methods for providing insights on data of various platforms are provided. One or more requests for insights on data of at least one of an automation platform, a process mining platform, or a task mining platform are received. In response to receiving the one or more requests, a response comprising the insights is generated using a language model that was trained using training data of one or more automation platforms, one or more process mining platforms, and one or more task mining platforms. The response is output.