On-Device Emoji Classifier Trained with GPT-based Data Augmentation for a Mobile Keyboard

2024-11-06

arXiv.org

AI Summary

This work addresses real-time emoji prediction for mobile keyboards under four key challenges: long-tailed class distribution, severe label imbalance, stringent on-device resource constraints, and the need for user personalization. Method: We propose a lightweight on-device emoji classifier featuring (1) a novel GPT-3.5–driven, human-free data augmentation paradigm that synthesizes diverse sentences and fine-grained emoji labels to enrich rare-class samples; (2) unsupervised label mapping coupled with temporal interpolation of user historical behavior to enable preference-aware inference; and (3) a quantization-ready architecture built upon MobileBERT for efficient edge deployment. Contribution/Results: Deployed in SwiftKey, our model significantly improves rare-emoji prediction accuracy, increases emoji click-through rate by 12.7%, achieves end-to-end latency under 80 ms, and consumes less than 15 MB memory—demonstrating strong trade-offs among coverage, fairness, and personalization.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Humans and AI: Intelligent User Interfaces

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Emojis improve communication quality among smart-phone users that use mobile keyboards to exchange text. To predict emojis for users based on input text, we should consider the on-device low memory and time constraints, ensure that the on-device emoji classifier covers a wide range of emoji classes even though the emoji dataset is typically imbalanced, and adapt the emoji classifier output to user favorites. This paper proposes an on-device emoji classifier based on MobileBert with reasonable memory and latency requirements for SwiftKey. To account for the data imbalance, we utilize the widely used GPT to generate one or more tags for each emoji class. For each emoji and corresponding tags, we merge the original set with GPT-generated sentences and label them with this emoji without human intervention to alleviate the data imbalance. At inference time, we interpolate the emoji output with the user history for emojis for better emoji classifications. Results show that the proposed on-device emoji classifier deployed for SwiftKey increases the accuracy performance of emoji prediction particularly on rare emojis and emoji engagement.

Problem

• On-device emoji prediction
• Data imbalance in emoji datasets
• User-specific emoji classification

Innovation

• On-device emoji classifier
• GPT-based data augmentation
• User history interpolation