Gesture Typing Accuracy via Contextual Word Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Gesture typing in messaging and social media applications is prone to inaccuracies due to user imprecision, leading to 'input noise' and incorrect word suggestions, particularly with issues like confounding words, double letters, and uniform swiping speed, which results in users often resorting to manual typing.
Innovation Solution
The system employs contextual data analysis using natural language processing, lexical chaining, and semantic understanding to generate a high-priority list of word suggestions based on user gestures, blending this with the lexicon to improve accuracy and relevance, and adjusts probabilities using context-specific databases and user history.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If gesture typing uses finger trajectory analysis to suggest words, then text entry speed is improved, but word suggestion accuracy deteriorates due to user imprecision and input noise
Solution Approach 1:
The patent introduces contextual data as an intermediary factor between the finger trajectory and word selection. Instead of directly mapping trajectory to word based solely on lexical probability, the system incorporates contextual information (conversation history, topic, user preferences) as a mediator that refines the word suggestion, thereby compensating for trajectory imprecision while maintaining fast input speed
Solution Approach 2:
The system changes the parameter used for word ranking from solely lexical probability (based on trajectory) to a composite parameter that weights both lexical probability and contextual relevance. This parameter transformation allows the system to maintain speed while improving accuracy by adjusting the ranking criteria dynamically based on context
2Speed
If the system suggests words based on highest lexical probability match, then processing speed is improved, but reliability deteriorates due to confounding words with similar trajectories
Solution Approach 1:
The system incorporates feedback from contextual analysis to refine word suggestions. By continuously monitoring the conversation context and comparing it against suggested words, the system can adjust and refine its recommendations, rejecting confounding words that don't align with contextual patterns while maintaining fast processing through efficient contextual filtering
3Difficulty of detecting and measuring
If the system analyzes pause duration to detect double letters, then detection capability is improved, but measurement precision deteriorates due to user imprecision in pausing
Solution Approach 1:
The patent uses contextual data as an intermediary to validate detected double letters. Instead of relying solely on pause duration thresholds, the system checks whether the detected double letter makes sense in the current contextual sequence, thereby compensating for imprecise pause timing measurements while maintaining enhanced detection capability
Data Source
AI summary
Systems and methods for improving gesture-based typing in applications where contextual data is used to provide improved word suggestions corresponding to a gesture (i.e., trajectory of hand movement on a digital keyboard) are described. A high-level context is determined by applying techniques such as natural language processing to one or more data categories. Based on the determined high-level context, context specific databases are accessed to obtain additional context specific words. A high priority list (HPL) is generated based on the both the obtain context specific words and the lexicon. The words in the HPL are ranked based on their probability of matching the gesture entered by the user and a top ranked words or words are displayed for input. If suggested word is not accepted by the user, then based on determining a similarity between two sequential gestures, previously suggested words are not repeatedly suggested.