Visual Exploration Framework for Linguistic Expression Modification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current predictive models are unintelligible to human users, making them difficult to understand and modify, which hinders optimization and efficiency in applications such as product identification and spam email filtering.
Innovation Solution
A computer-implemented method using a visual exploration framework to receive and modify linguistic expressions, allowing users to visually present, sort, filter, and select expressions for inclusion in classification models, enabling human interaction and improvement of model rules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If predictive models are deployed to achieve high accuracy in classification tasks, then productivity is improved, but the models become unintelligible and difficult to modify
Solution Approach 1:
The patent segments the predictive model into discrete linguistic expressions (LEs) that can be individually examined and modified. Each LE represents a specific classification rule that can be independently manipulated, allowing users to understand and adjust model behavior without dealing with the complexity of the entire neural network.
Solution Approach 2:
The patent introduces an intermediary layer between the neural network and the user interface. This intermediary translates the neural network's internal representations into human-readable linguistic expressions, serving as a mediator that makes the model's decision logic accessible to users while maintaining the underlying neural network's computational power.
2Adaptability or versatility
If predictive models are made more complex to handle diverse classification tasks, then adaptability is improved, but the models become harder to understand and modify
Solution Approach 1:
The patent breaks down the complex model structure into discrete, manageable linguistic expressions. Each expression corresponds to a specific classification rule that can be independently modified, making it easier to adapt the model to different tasks without redesigning the entire architecture.
Solution Approach 2:
The patent enables dynamic modification of the model by allowing users to add, remove, or modify linguistic expressions interactively. This dynamic approach allows the model to adapt to new classification tasks by simply updating the linguistic expressions rather than retraining the entire neural network.
3Ease of operation
If additional computing resources are allocated to interact with predictive models, then ease of operation is improved, but the cost and complexity of the system increases
Solution Approach 1:
The patent introduces a visual exploration framework as an intermediary that simplifies user interaction with the model. This framework provides user-friendly interfaces for examining and modifying linguistic expressions, reducing the need for complex computing resources while maintaining ease of operation.
Solution Approach 2:
The patent creates a simplified representation of the model in the form of linguistic expressions that can be easily manipulated. This copying approach allows users to interact with a simplified version of the model logic without needing to directly manipulate the complex underlying neural network, reducing system complexity while maintaining ease of operation.
Data Source
AI summary
A computer-implemented method according to one embodiment includes receiving a plurality of linguistic expressions (LEs); changing one or more conditions of the plurality of linguistic expressions to create an updated plurality of linguistic expressions, utilizing a visual exploration framework (VEF) that visually presents to a user each of the plurality of linguistic expressions; and including the updated plurality of linguistic expressions in a model used to classify input sentences. According to another embodiment, a computer-implemented method includes receiving (i) a set of linguistic expressions (LEs) and (ii) a set of labeled data as input, where the LEs are logical combinations of predicates learned from the labeled data, and each data point in the labeled data comprises a piece of text and ground-truth labels; presenting the LEs in a visual exploration framework; and allowing a user to sort, filter, subset, and select LEs based on different criteria, utilizing the framework.