Context-Aware String Prediction Buffer for Mobile Typing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mobile device typing methods, such as soft input panels, fail to provide accurate word predictions by not considering recently used words or phrases in user conversations, leading to frustrating typing experiences.
Innovation Solution
A mobile device system that stores strings from current threads in a buffer and uses them to generate predictions, adjusting probabilities of dictionary-based predictions based on thread frequency, to provide contextually relevant suggestions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If word predictions are generated using only dictionary matching and general population frequency, then the prediction system is simple and fast, but the predictions are not contextually accurate and do not reflect user-specific usage patterns
Solution Approach 1:
The system performs preliminary actions by storing strings from current threads in a buffer before prediction is needed. This pre-processing of contextual data allows the prediction system to access relevant user-specific strings quickly when making predictions, improving accuracy without adding complexity during the actual typing interaction
Solution Approach 2:
The patent introduces a buffer as an intermediary component that stores contextual strings from current threads. This buffer acts as a mediator between the dictionary-based prediction system and the user-specific contextual information, allowing the system to incorporate user-specific patterns without fundamentally changing the underlying prediction algorithm
2Measurement precision
If the system incorporates user-specific and context-aware predictions by analyzing recent conversations, then prediction accuracy improves, but the processing time and computational resources increase
Solution Approach 1:
By pre-storing strings from current threads in a buffer, the system prepares contextual data in advance so that during typing, it can quickly retrieve and compare against entered characters without performing complex analysis in real-time, thus maintaining fast response while improving accuracy
Solution Approach 2:
The system performs partial analysis by only considering strings from the current thread buffer that match the prefix of entered characters, rather than analyzing all possible strings. This selective approach reduces processing time while still capturing the necessary contextual information for accurate predictions
3Productivity
If the system uses a soft input panel with word predictions, then typing speed may improve, but user frustration increases when predictions do not match the user's intended words
Solution Approach 1:
The system applies local quality by tailoring predictions to the specific local context of the current thread. Instead of providing generic predictions based on overall language frequency, it prioritizes strings from the current conversation thread, making predictions locally relevant to what the user is actually typing about, thus improving both speed and user experience
Solution Approach 2:
The system incorporates feedback from the current thread strings that have been previously used by the user. By analyzing these past inputs in the context and using them to inform future predictions, the system learns from user behavior patterns and continuously improves prediction relevance, reducing frustration and enhancing ease of operation
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In a mobile device, the strings that make up a current thread in application are stored in a buffer. A thread is a series of communications sent between a user of the mobile device and one or more other users, such as a series of text messages or a series of emails. When the user enters text to continue the current thread, characters of the entered text are compared against the strings stored in the buffer, and strings that have the entered text as a prefix are selected and displayed to the user as string predictions in a soft input panel. Alternatively, the strings in the buffer may be used to increase the probabilities of matching strings in a dictionary or model that is used to select string predictions.