Towards understanding evolution of science through language model series

2024-09-15

arXiv.org

AI Summary

This study addresses the challenge of modeling temporal evolution in scientific text. To this end, we propose the AnnualBERT family of models: RoBERTa-based architectures employing whole-word tokenization and a year-by-year incremental pretraining paradigm from scratch, trained on 1.7 million arXiv papers (up to 2008) to construct both a foundational base model and annually updated variants. For the first time, language models are explicitly aligned with the temporal granularity of scientific advancement—departing from subword tokenization and monolithic model paradigms—and we design quantifiable probing tasks to analyze dynamic concept representation shifts and knowledge forgetting. AnnualBERT achieves state-of-the-art performance on domain-specific NLP and arXiv citation link prediction tasks; matches mainstream models on general NLP benchmarks; and reveals empirically grounded patterns of semantic decay and reconstruction in scientific concepts over time.

Technology Category

Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal ReasoningCognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologiesWeb Mining and Content Analysis: Models for Web evolutionGraph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

We introduce AnnualBERT, a series of language models designed specifically to capture the temporal evolution of scientific text. Deviating from the prevailing paradigms of subword tokenizations and"one model to rule them all", AnnualBERT adopts whole words as tokens and is composed of a base RoBERTa model pretrained from scratch on the full-text of 1.7 million arXiv papers published until 2008 and a collection of progressively trained models on arXiv papers at an annual basis. We demonstrate the effectiveness of AnnualBERT models by showing that they not only have comparable performances in standard tasks but also achieve state-of-the-art performances on domain-specific NLP tasks as well as link prediction tasks in the arXiv citation network. We then utilize probing tasks to quantify the models' behavior in terms of representation learning and forgetting as time progresses. Our approach enables the pretrained models to not only improve performances on scientific text processing tasks but also to provide insights into the development of scientific discourse over time. The series of the models is available at https://huggingface.co/jd445/AnnualBERTs.

Problem

• Capturing temporal evolution of scientific text
• Improving domain-specific NLP task performance
• Analyzing scientific discourse development over time

Innovation

• Uses whole words as tokens instead of subwords
• Base model pretrained on 1.7 million arXiv papers
• Progressively trained annually for temporal evolution