Smart Reply Generation Using Image Embedding Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for generating smart replies involving image files face challenges such as high computational resource usage and inefficient use of display space, particularly on mobile devices, due to the need to evaluate numerous parameters and display irrelevant suggestions.
Innovation Solution
A method that detects image files in messages, generates smart replies based on embedding vectors derived from query text, image data, and textual metadata, and displays these as selectable user interface elements, allowing users to select and transmit relevant replies or trigger image file searches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the computer system evaluates a vast number of parameters to provide relevant suggested replies, then the relevance and usefulness of smart replies is improved, but the computational resource usage increases
Solution Approach 1:
The patent segments the smart reply generation process into multiple stages: initial generation of candidate replies, evaluation against multiple parameters, filtering to remove irrelevant options, and selection of final replies. This segmentation allows the system to process vast numbers of parameters efficiently by breaking down the complex evaluation into manageable steps, maintaining high relevance while controlling computational resource usage.
Solution Approach 2:
The system applies different evaluation criteria and parameter weights depending on the specific context of each message and user situation. Rather than uniformly evaluating all parameters for all replies, the system adapts the evaluation process to prioritize relevant parameters locally, improving reply relevance while reducing unnecessary computational expenditure on irrelevant parameters.
2Adaptability or versatility
If the computer system displays an excessive number of smart replies, then the user is provided with more options, but the display space is wasted on irrelevant suggestions
Solution Approach 1:
The system dynamically adjusts the number of smart replies displayed based on available display space, user preferences, and message context. Rather than displaying a fixed number of replies, the system adaptively determines the optimal number to show, ensuring that display space is utilized efficiently while still providing users with sufficient reply options. This dynamic adjustment resolves the contradiction between providing versatility and optimizing display area usage.
3Speed
If the computer system generates and displays smart replies quickly, then the user experience is improved, but the computational load increases
Solution Approach 1:
The system performs preliminary processing and pre-computation of potential smart replies based on common message patterns and user behavior. By preparing candidate replies in advance and caching frequently used response templates, the system can quickly generate relevant replies when needed, reducing both the perceived response time and the instantaneous computational load during actual messaging interactions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Techniques for generating smart replies involving image files are disclosed herein. In some example embodiments, a computer system detects that a first message comprising a first image file has been transmitted from a first computing device of a first user to a second computing device of a second user, and generates a first plurality of smart replies based on a first embedding vector of the first image file, where the first embedding vector of the first image file is based on at least one of first textual metadata of the first image file, first image data of the first image file, and a first set of query text used by a first set of users in a first set of searches that resulted in the first image file being included in a first set of transmitted messages.