Conversation Trust Scoring With Vocal and Text Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies lack a standardized method to quantify and improve trust in conversations, relying heavily on subjective ratings, making it difficult to compare trust delivery among individuals/teams and track its improvement over time.
Innovation Solution
A system and method using machine learning models to objectively quantify and improve trust in conversations by analyzing vocal and textual features, employing machine-learning networks with tunable hyperparameters, and generating trust scores through cross-validation training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subjective ratings are used to assess trust, then trust assessment can be performed, but measurement precision and objectivity are poor
Solution Approach 1:
The patent replaces subjective human rating mechanisms with an automated machine learning system that processes vocal and text features. The ML model objectively calculates trust scores by analyzing acoustic properties (pitch, energy, zero-crossing rate) and textual characteristics (sentiment, word choice), eliminating human bias and improving measurement precision without requiring complex manual assessment procedures
Solution Approach 2:
The patent introduces vocal features and text features as intermediary variables between the conversation and the trust score. These features serve as measurable proxies for trust, allowing the system to quantify subjective trust concepts through objective acoustic and linguistic measurements that can be processed algorithmically
2Measurement precision
If machine learning models with multiple hyperparameters are used, then trust scoring accuracy is improved, but device complexity and training requirements increase
Solution Approach 1:
The patent performs preliminary actions by pre-processing vocal and text features before feeding them to the ML model. Vocal features such as pitch, energy, and zero-crossing rate are extracted and normalized in advance, as are textual features like sentiment scores and word embeddings. This pre-processing reduces the complexity of the main modeling task and enables more accurate trust scoring
Solution Approach 2:
The patent employs dynamic hyperparameter tuning through cross-validation to optimize model performance. The system adaptively selects the best combination of hyperparameters based on validation performance, allowing the model complexity to be dynamically adjusted to achieve optimal accuracy while avoiding unnecessary complexity
3Reliability
If cross-validation and hyperparameter tuning are performed, then model reliability is improved, but loss of time increases
Solution Approach 1:
The patent implements a balanced approach to cross-validation by performing k-fold validation with a moderate number of folds and iterations. This provides sufficient model reliability assessment without requiring exhaustive validation that would consume excessive time. The system finds an optimal balance between validation thoroughness and training time efficiency
Data Source
AI summary
The present disclosure relates to a system, a method, and a product for using machine learning models to quantify and/or improve trust in conversations. The system includes a non-transitory memory; and a processor in communication with the non-transitory memory. The processor executes the instructions to cause the system to: obtain a set of vocal features and a set of text features for each sample in audio samples; obtain a trust score for each sample; perform a preprocess to obtain a set of input features for each sample; determine a type of machine-learning algorithm for the machine-learning network; tune a set of hyper parameters for the machine-learning network; generate a predicated trust score by the machine-learning network with the sets of input features for each sample; and train the machine-learning network based on the predicated trust score and the trust score for each sample to obtain the training result.