Enhancing SDG-Text Classification with Combinatorial Fusion Analysis and Generative AI

2026-02-13 · Source: cs.CL updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, extended

Summary

This paper introduces a method to enhance text classification for the UN's 17 Sustainable Development Goals (SDGs) by combining multiple machine learning models using Combinatorial Fusion Analysis (CFA). The approach addresses challenges like interconnected SDG concepts and data scarcity by employing generative AI, specifically ChatGPT, to create synthetic training data. Five base models were used: SDG Classy (LDA-based), LinkedSDG (semantic web), SDG Mapper (keyword-based), Convolutional Neural Network (CNN), and Random Forest. The CFA technique, which leverages rank-score characteristic functions and cognitive diversity, achieved an average precision@1 of 96.73% on a 306-document test set, outperforming the best individual model (Model A at 95.42%) and a fine-tuned BERT-base model (94.46%). The study also highlights how CFA can complement human expert judgment and mitigate risks associated with noisy synthetic data.

Key takeaway

For NLP engineers developing text classification systems for complex, interconnected categories like the UN SDGs, adopting a Combinatorial Fusion Analysis (CFA) approach is highly effective. You should consider generating synthetic training data using large language models like ChatGPT to address data scarcity, and then combine multiple diverse base models with CFA to achieve superior accuracy and robustness, potentially surpassing even fine-tuned transformer models. This strategy can also help mitigate the impact of noisy synthetic data.

Key insights

Combinatorial Fusion Analysis (CFA) significantly enhances SDG text classification by combining diverse models and synthetic data.

Principles

• Model fusion improves performance over single optimized models.
• Cognitive diversity quantifies dissimilarity between scoring systems.
• Generative AI can augment scarce labeled training data.

Method

The method involves generating synthetic training data with ChatGPT, training diverse base classifiers, and then combining their outputs using CFA's rank-score and cognitive diversity functions to achieve superior classification accuracy.

In practice

• Use ChatGPT for synthetic data generation to overcome data scarcity.
• Apply CFA to combine multiple classifiers for improved accuracy.
• Consider multi-label classification for interconnected SDG concepts.

Topics

• SDG Text Classification
• Combinatorial Fusion Analysis
• Generative AI
• Model Fusion
• Natural Language Processing

Code references

• SeaCelo/SDG-samples
• dialogicnl/eur-sdg

Best for: NLP Engineer, Research Scientist, AI Researcher, AI Scientist, Machine Learning Engineer

Related on AIssential

• Learning Perspectivist Social Meaning via Demographic-Conditioned Fusion Embeddings — Demographic-conditioned fusion embeddings significantly improve modeling of perspectivist social meaning in language.
• MASF: A Multi-Model Adaptive Selection Framework for Abstractive Text summarization — Combining multiple fine-tuned summarization models with adaptive selection improves output quality and robustness.
• Grammar as an Injectable: A Trojan Horse to NLP | Towards Data Science — CCG offers a non-statistical, algebraic framework for NLP, treating words as functions to ensure grammatical consistency and enhance interpretability.
• On GATE, Text and Social Media Analysis, and Detecting Misinformation Online — A multi-layer analytical pipeline effectively extracts and categorizes digital identity aspects of agrifood products from online discourse.
• SenFlow: Inter-Sentence Flow Modeling for AI-Generated Text Detection in Hybrid Documents — SenFlow improves AI-generated text detection in hybrid documents by modeling inter-sentence dependencies and using a new benchmark.
• NLP-based Page Classification for Efficient LLM Extraction from Brazilian Public Tender Documents — Combining NLP page classification with LLM extraction significantly reduces token count and cost for document processing.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CL updates on arXiv.org.