Application-Specific Text Suggestions Using Combined App Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems generate the same text suggestions across different applications on electronic devices, which can lead to user dissatisfaction due to context mismatch, affecting the user experience.
Innovation Solution
A machine-learning model is trained to generate application-specific text suggestions by combining user input embeddings with application name embeddings, using algorithms like CHAR-CNN and one-hot encoding, and a natural language processing model such as LSTM or RWKV, to provide contextually relevant suggestions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the electronic device generates the same text suggest in response to a user textual input across all applications, then the system complexity is reduced and ease of operation is improved, but the adaptability to different application contexts deteriorates and user satisfaction decreases
Solution Approach 1:
The system segments the text suggestion generation process by application context. It creates separate processing paths for different applications (messaging, social network, map, etc.) by generating application-specific embeddings and combining them with input embeddings to produce contextually appropriate suggestions for each application type.
Solution Approach 2:
The system applies local quality by making text suggestions application-specific rather than uniform across all applications. Each application receives tailored suggestions based on its contextual characteristics, with the NLPM generating different suggestion qualities and types suited to each application's purpose.
2Adaptability or versatility
If the electronic device generates application-specific text suggestions using a trained machine-learning model, then the adaptability to different application contexts is improved, but the device complexity increases
Solution Approach 1:
The system uses a universal Natural Language Processing Model (NLPM) that serves multiple functions across different applications. The single NLPM handles text suggestion generation for various application types (messaging, social network, map, etc.) by processing application-specific embeddings, eliminating the need for separate models for each application.
Solution Approach 2:
The system introduces application name embeddings as an intermediary layer between the application context and the NLPM. These embeddings serve as mediators that translate diverse application contexts into a unified representation that the NLPM can process, simplifying the overall system architecture.
3Reliability
If the electronic device provides contextually relevant text suggestions tailored to each application, then user satisfaction is improved, but the loss of information about application context increases without proper embedding generation
Solution Approach 1:
The system performs preliminary action by generating application name embeddings in advance before text suggestion generation. This pre-processing step captures and preserves application context information in embedding form, ensuring that the NLPM receives complete contextual information needed for reliable suggestion generation.
Solution Approach 2:
The system creates composite information representations by combining application name embeddings with input text embeddings. This composite embedding contains both application context and user input information, preserving all necessary information for generating contextually relevant suggestions without information loss.
Data Source
AI summary
A method and an electronic device for generating text suggests for texts input in applications executed on the electronic device are provided. The method comprises: receiving a textual user input; generating a first vector embedding representative of the textual user input; generating a second vector embedding representative of an application name of a given application of the plurality of applications, to which the textual user input has been made; combining the first and second vector embeddings to generate a combined vector embedding for the textual user input; feeding the combined vector embedding to a natural language processing model (NLPM) to generate a text suggest for the user to select as input, to the given application, following the textual user input; and outputting the text suggest to enable the user of the electronic device to input the text suggest after the textual user input to the given application.