Adaptive Text Prediction Weighting for Input Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text input systems for electronic devices rely solely on recency or frequency for ordering predictions, which can lead to user inefficiency as they require scrolling through lists or entering additional characters to find intended terms, lacking relevance based on high-level characteristics like topic, genre, or authorship.
Innovation Solution
A system employing machine learning techniques, specifically classification, to make real-time category predictions for user-entered text, reordering and weighting text predictions to prioritize relevant options, using a classifier trained on pre-labelled text data to generate category-weighted predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If text predictions are ordered solely on the basis of recency or frequency, then the system is simple to implement, but user efficiency deteriorates as users must scroll through lists or enter additional characters to find intended terms
Solution Approach 1:
The patent changes the ordering parameters of text predictions from simple recency or frequency metrics to a multi-dimensional classification system that incorporates topic, genre, and authorship characteristics. This allows the system to prioritize predictions based on their relevance to the current textual context, significantly improving user efficiency without requiring users to scroll through lengthy lists.
Solution Approach 2:
The patent introduces additional dimensions for organizing text predictions by creating a hierarchical classification system with multiple levels (broad categories, sub-categories, and fine-grained classifications). This dimensional expansion allows the system to present predictions in a more meaningful order based on their semantic relationship to the input text, resolving the contradiction between simplicity and user efficiency.
2Reliability
If the system uses basic dictionary with usage frequency statistics, then the system requires minimal memory, but prediction accuracy deteriorates lacking high-level characteristics like topic, genre, or authorship
Solution Approach 1:
The patent segments the prediction system into multiple hierarchical levels: broad category predictions, sub-category predictions, and fine-grained term predictions. Each level processes and filters predictions independently, allowing the system to incorporate rich contextual information (topic, genre, authorship) while managing memory usage through progressive refinement rather than loading all data at once.
Solution Approach 2:
The patent performs preliminary classification of predictions into categories and sub-categories before presenting the final ordered list to the user. This preliminary action allows the system to pre-process and organize predictions based on high-level characteristics, improving prediction accuracy by filtering and prioritizing terms that are more likely to be relevant to the current textual context.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates generally to a system and method for text prediction and, in particular, a system and method for the adaptive weighting of text predictions. The present invention provides a modular language model-based text prediction system for the adaptive weighting of text prediction components and a method for doing so. The system uses category predictions, generated by a timed aggregated perceptron classifier, to weight, using a weighting module, text predictions generated by a text prediction engine comprising at least one predictor.