Hierarchical Probabilistic Graph for Mental State Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to accurately model and infer the mental state of users, particularly in virtual agent interactions, as they struggle to simultaneously determine and track personality, mood, and emotion from text and speech communications effectively.
Innovation Solution
The use of a specialized language processor that employs a hierarchical probabilistic graphical model to analyze communications, maintaining separate states for personality, mood, and emotion, and outputs the mental state through a graphic user interface or to another process, with edges connecting nodes and probability values to determine the highest probability path.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing systems attempt to model user mental state, then personalization capability is improved, but accuracy in simultaneously determining personality, mood, and emotion deteriorates
Solution Approach 1:
The patent segments the mental state modeling into three distinct hierarchical levels: personality traits (stable, long-term characteristics), mood (temporary affective state), and emotion (momentary emotional reactions). Each level is modeled separately with its own probabilistic variables and inference mechanisms, allowing accurate simultaneous determination of all three components through the directed probabilistic graphical model.
2Loss of information
If a comprehensive model tracks multiple mental state variables, then customer modeling depth is improved, but system complexity increases
Solution Approach 1:
The patent implements a nested hierarchical structure where personality traits form the outermost stable layer, mood represents the intermediate temporary state layer, and emotion constitutes the innermost momentary reaction layer. This nested organization allows the system to track multiple mental state variables comprehensively while managing complexity through hierarchical abstraction, with each level influencing but being distinct from the others.
Solution Approach 2:
The patent adds temporal dimensionality to the mental state model by distinguishing between stable (personality), temporary (mood), and momentary (emotion) states. This dimensional differentiation allows comprehensive tracking of mental state variables across different time scales, transforming a complex simultaneous modeling problem into a structured multi-dimensional framework that is more manageable.
3Speed
If real-time analysis of communications is performed, then responsiveness is improved, but computational resources required increase
Solution Approach 1:
The patent performs preliminary action by pre-defining the hierarchical probabilistic graphical model structure with all necessary nodes (personality, mood, emotion variables) and edges (probabilistic relationships) before real-time analysis begins. During real-time communication analysis, the system only needs to input observed data into the pre-structured model and compute the highest probability path, significantly reducing computational resources compared to building and optimizing the model during real-time processing.
Solution Approach 2:
The patent uses copying by representing the complex mental state inference through a simplified directed probabilistic graphical model that copies the essential relationships between mental state variables. This graphical model copy allows efficient real-time computation by replacing complex probabilistic reasoning with graph path finding, where the highest probability path through the pre-constructed graph provides the inferred mental state without requiring heavy computational resources during real-time operation.
Data Source
AI summary
Methods and systems monitor communications between users and analyze the communications to simultaneously determine, for a current time period, mental state variables of one of the users. Such mental state variables include the emotion of the user, the mood of the user, and the personality of the user. Additionally, such methods aggregate the emotion, the mood, and the personality using a hierarchical probabilistic graphical model that determines the highest probability path through a directed probabilistic graph to infer the mental state of the user. The directed probabilistic graph maintains a single state for the personality for the time period, and maintains multiple states for the emotion and the mood for the time period. These methods and systems output the mental state of the user.