Smart Text Span Selection on Touch Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current text selection methods on touch-enabled devices, such as smartphones and tablets, offer lower selection acuity compared to traditional pointing devices, making it cumbersome to select multi-word text spans accurately and contextually relevant to the document content.
Innovation Solution
The smart selection technique employs a machine-learned ensemble model and hyperlink intent model to predict the intended text span by re-scoring candidate text spans generated from various methods, using a dataset of true user selections and leveraging linguistic unit detectors and heuristic methods to provide accurate multi-word text span predictions with a single touch gesture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional pointing devices are used for text selection, then selection precision is improved, but device portability and ease of operation deteriorate
Solution Approach 1:
The patent replaces mechanical pointing devices (mouse, trackpad) with a touch-based interface that uses software algorithms to determine text span selections. The system substitutes physical mechanical selection with capacitive touch gestures combined with machine learning models that predict user intent based on contextual features.
Solution Approach 2:
The system enables the touch interface to automatically determine the intended text span without requiring precise manual selection. The machine learning model autonomously analyzes contextual features and predicts the user's intended selection, making the interface self-correcting and reducing the need for high-precision manual input.
2Ease of operation
If touch gestures are used for text selection, then ease of operation is improved, but selection precision deteriorates
Solution Approach 1:
The system incorporates feedback loops where the machine learning model continuously refines its predictions based on user interactions and contextual analysis. The model provides feedback by suggesting text spans and adjusting predictions based on user confirmation or correction patterns, improving accuracy over time while maintaining ease of use.
Solution Approach 2:
The patent changes the parameters of text span selection from direct coordinate-based selection to probabilistic predictions based on multiple contextual features. The system transforms the selection problem from a spatial precision task to a contextual inference task, using features like text semantics, user behavior patterns, and document structure to determine the intended span.
3Measurement precision
If machine learning models are used to predict text spans, then selection accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent segments the text span prediction task into multiple independent feature extraction components. The machine learning system divides the complex prediction problem into separate analyses of textual, contextual, and behavioral features, processing each segment independently before combining results. This modular approach reduces computational complexity while maintaining prediction accuracy.
Solution Approach 2:
The system performs partial analysis by focusing on the most discriminative features for text span prediction rather than analyzing all possible document characteristics. The machine learning model selectively processes only the relevant contextual features needed for accurate prediction, avoiding unnecessary computational overhead from exhaustive analysis.
Data Source
AI summary
A text span forming either a single word or a series of two or more words that a user intended to select is predicted. A document and a location pointer that indicates a particular location in the document are received and input to different candidate text span generation methods. A ranked list of one or more scored candidate text spans is received from each of the different candidate text span generation methods. A machine-learned ensemble model is used to re-score each of the scored candidate text spans that is received from each of the different candidate text span generation methods. The ensemble model is trained using a machine learning method and features from a dataset of true intended user text span selections. A ranked list of re-scored candidate text spans is received from the ensemble model.