Taxonomy Maintenance In The Wild Over Evolving Scholarly Data: Reliability, Efficiency, and Cost-Effectiveness

2026-07-10

AI Summary

This work addresses the challenge of rapidly outdated taxonomies in the face of accelerating academic knowledge growth by proposing GIST, a framework for continuously maintaining research topic classifications in dynamic document repositories such as arXiv. GIST uniquely integrates expert evidence guidance, geometric structure induction, and bidirectional embedding alignment to construct an incrementally updatable global taxonomy in a box embedding space, leveraging local hierarchical structures extracted from “Related Work” sections. It further incorporates a novelty-aware core-set selection strategy and a hypothesized concept generation mechanism. Evaluated on real-world arXiv data, GIST achieves relative improvements of 11.0% and 13.1% in Node F1 and Edge F1 scores, respectively, while requiring only 9.6% of the runtime and 12.7% of the cost of the strongest baseline, thereby significantly balancing accuracy, efficiency, and scalability.

Abstract

The rapid growth of scientific publications makes scholarly taxonomies quickly obsolete. We study taxonomy maintenance in the wild, a new problem that moves beyond static construction by continuously adapting taxonomies to evolving scholarly repositories, such as arXiv, for a given research topic. We propose GIST, a robust framework for maintaining evolving taxonomies. Unlike purely LLM-centric approaches, GIST grounds structure induction in expert-curated evidence by extracting partial hierarchies from the "Related Work" sections of papers. It integrates these partial taxonomies into a unified global taxonomy in a geometric box-embedding space, where box containment encodes the inductive bias of is-a relations. To connect semantics with geometric structure, GIST learns a bidirectional mapping between word embeddings and box embeddings. For efficient incremental updates, GIST uses novelty-aware coreset selection to update the model with representative historical signals and new evidence, avoiding costly full retraining. To handle high-velocity paper streams under user-specific token budgets, GIST further combines a hypothesized concept generator with a cost-effective evidence retrieval module. Experiments on real-world arXiv datasets show that GIST outperforms state-of-the-art baselines, improving Node F1 and Edge F1 by 11.0% and 13.1% over the strongest baseline while requiring only 9.6% of its runtime and 12.7% of its monetary cost.

Problem

• taxonomy maintenance
• evolving scholarly data
• academic taxonomy
• dynamic classification
• knowledge organization

Innovation

• taxonomy maintenance
• box embedding
• incremental learning
• evidence grounding
• cost-effective AI