Natural Language Planning for Accurate Skill Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Software tools are underutilized due to lack of education, awareness, unintuitive user interfaces, and difficulty in finding the right tool among a vast selection, particularly affecting security analysts who are unaware of available tools and their optimal usage.
Innovation Solution
A machine learning model generates a plan using natural language prompts to select and order skills based on user intent, leveraging embedding representations and meta-prompts to facilitate seamless tool integration and operation, allowing users to interact conversationally.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a large selection of software tools is provided to security analysts, then the capability and versatility of the system is improved, but the difficulty of finding and utilizing the right tool increases
Solution Approach 1:
The patent introduces an intermediary system consisting of embedding models and plan generators that automatically match user natural language requests with appropriate software tools. This intermediary layer eliminates the need for users to manually search through vast tool selections, as the system autonomously identifies and sequences the required tools based on semantic understanding of the request.
Solution Approach 2:
The system enables self-service by allowing users to interact with the tool ecosystem through natural language without requiring knowledge of available tools or their intricacies. The embedding model and plan generator automatically handle tool selection and sequencing, making the system serve itself by matching requests to tools without human intervention in the selection process.
2Measurement precision
If comprehensive tool knowledge is required for effective utilization, then the precision of tool selection is improved, but the complexity of user education and awareness increases
Solution Approach 1:
The patent replaces the mechanical system of user education and manual tool selection with an automated semantic matching system. Instead of requiring users to learn about tools through traditional education methods, the system uses embedding models to computationally match natural language requests with appropriate tools, substituting human learning with machine learning.
Solution Approach 2:
The system changes the parameter of tool selection from requiring user knowledge state to using semantic embedding space. By transforming the selection criterion from human-understandable tool descriptions to mathematical embedding representations, the system achieves accurate matching without requiring users to understand the underlying complexity.
3Measurement precision
If traditional user interfaces are used for tool selection, then the precision of command execution is improved, but the ease of operation deteriorates
Solution Approach 1:
The patent inverts the traditional interface paradigm by allowing users to speak naturally about their goals rather than requiring them to learn specific commands or syntax. Instead of the user adapting to the system's interface requirements, the system adapts to the user's natural language, generating appropriate tool sequences from conversational input.
Data Source
AI summary
Machine learning models are used to generate a plan that responds to a user request. The plan includes one or more skills selected from a list of available skills. The prompt may be written in natural language, enabling the user to express their intent without having to know which skills are available or their intricacies. In some configurations, a skill is included in the plan if an embedding representation of an example prompt associated with the skill is within a defined distance of an embedding representation of the user request. Additionally, or alternatively, the embedding distance computations are used to narrow the list of available skills, which is then used to construct a meta-prompt that selects a skill. Skills listed in the meta-prompt may include data types of parameters and return values. This allows the model that processes the meta-prompt to order skills based on data type compatibility.