Compact Model for Theme-Based Font Search Latency Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current font selection processes are tedious and inefficient, requiring users to search through extensive lists of fonts, making it difficult to find suitable fonts based on theme or intent.
Innovation Solution
The use of a compact model built using a pre-trained large language model to provide a theme-based font search, which generates font tag embeddings and query embeddings to suggest relevant font categories based on user input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a traditional large language model is used for font search, then the model can understand theme and intent, but the latency is high and model complexity is large
Solution Approach 1:
The patent segments the large language model into a compact model that retains only the essential embedding generation capability. Instead of using the full LLM for both understanding and search, the system extracts and deploys a specialized compact model that handles theme-based font search specifically, separating the general language understanding function from the specific search function.
Solution Approach 2:
The patent extracts the embedding generation capability from the pre-trained large language model and isolates it as a standalone compact model. This extracted component is then optimized for the specific task of font search, removing unnecessary complexity while preserving the thematic understanding capability.
2Measurement precision
If a traditional large language model is used for font search, then the model can generate accurate font suggestions, but the model complexity is high
Solution Approach 1:
The patent extracts only the necessary embedding generation functionality from the large language model, creating a compact model that eliminates unnecessary complexity. This extracted component maintains accuracy in understanding theme and intent while significantly reducing model size and computational requirements.
Solution Approach 2:
The patent replaces the expensive, complex large language model with a cheaper, simpler compact model that achieves the same functional goal for font search. The compact model is optimized specifically for this task, providing comparable accuracy with much lower resource requirements.
Data Source
AI summary
Aspects of the disclosure include methods and systems for providing a theme based font search in a manner that bypasses the latency limitations inherent to large language models and ensures domain-relevant answers. An exemplary method can include inputting a font category to a compact model to generate a font tag embedding and inputting a search query to the compact model to generate a query embedding. A distance measure is determined between the query embedding and the font tag embedding. The font category for the font tag embedding is provided as a suggested font category responsive to the distance measure satisfying a predetermined threshold.