ARISE: Iterative Rule Induction and Synthetic Data Generation for Text Classification

2025-02-09

AI Summary

To address weak rule interpretability and limited performance in few-shot and multilingual text classification, this paper proposes an iterative rule induction and synthetic data generation framework. Methodologically, it automatically induces generalized, interpretable rules via syntactic n-grams and establishes a rule-driven closed-loop iteration: rules provide supervision signals to refine synthetic data quality, while high-quality synthetic data, in turn, enables rule refinement; the framework supports bootstrapped rule selection and unified adaptation to both in-context learning (ICL) and fine-tuning. Its key innovation is the first realization of synergistic, closed-loop optimization between rule induction and data generation—without requiring external annotations. Experiments span three full-data, eight few-shot, and seven multilingual settings. Using either the induced rules or synthetic data alone yields substantial improvements over strong baselines—including complex contrastive learning—significantly enhancing performance in few-shot, multilingual, and in-context learning scenarios.

Technology Category

Natural Language Processing: Code Generation / Program Synthesis from Natural LanguageMachine Learning: Neuro-Symbolic LearningData Mining & Knowledge Management: Rule Mining & Pattern Mining

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingSemantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsEconomics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

We propose ARISE, a framework that iteratively induces rules and generates synthetic data for text classification. We combine synthetic data generation and automatic rule induction, via bootstrapping, to iteratively filter the generated rules and data. We induce rules via inductive generalisation of syntactic n-grams, enabling us to capture a complementary source of supervision. These rules alone lead to performance gains in both, in-context learning (ICL) and fine-tuning (FT) settings. Similarly, use of augmented data from ARISE alone improves the performance for a model, outperforming configurations that rely on complex methods like contrastive learning. Further, our extensive experiments on various datasets covering three full-shot, eight few-shot and seven multilingual variant settings demonstrate that the rules and data we generate lead to performance improvements across these diverse domains and languages.

Problem

• Iterative rule induction
• Synthetic data generation
• Text classification improvement

Innovation

• Iterative rule induction
• Synthetic data generation
• Bootstrapping for filtering