大型语言模型可以从分类学描述中提取形态数据,但它们的随机性质使自动化具有挑战性:对澳大利亚Asteraceae的测试
Alexander N Schmidt-Lebuhn1, Nunzio Knerr1
1CSIRO, Centre for Australian National Biodiversity Research (a joint venture of Parks Australia and CSIRO), Clunies Ross Street, Canberra ACT 2601, Australia Centre for Australian National Biodiversity Research Canberra Australia.
PhytoKeys
|August 29, 2025
概括
大型语言模型 (LLM) 可以从分类学文献中提取形态数据,但精确的指令对于准确性和可重复性至关重要. 开源模型表现不同,与专有模型相比,一致性较低.
科学领域:
- 植物学和分类学
- 计算生物学
- 生物信息学
背景情况:
- 形态数据对于分类学,进化生物学,生态学和物种识别至关重要.
- 不同于DNA序列或样本数据,缺乏对形态数据的集中数据库.
- 在分类学文献中,形态数据主要被"锁定",阻碍了可访问性.
研究的目的:
- 探索使用大型语言模型 (LLM) 和光学字符识别 (OCR) 来自动提取形态数据的可行性.
- 为澳大利亚本土和引入的Asteraceae填写一个分类群×字符表,其中包含51个形态字符.
- 从分类学描述中评估LLM在采矿形态数据中的准确性,可重复性和局限性.
主要方法:
- 使用ChatGPT 4o处理1121种澳大利亚星座的分类描述.
- 在945个种类中提取了51个形态特征的数据 (95个属,838个物种/亚种类).
- 通过视觉检查评估数据的准确性,并通过使用相同的开源LLM重复推断来评估可重复性.
主要成果:
- 在视觉检查中,缺失数据率为51.1%,整体错误率为5.8%.
- 错误率在cypsela和pappus字符中最低 (2.1%),其中常见的错误涉及胸膜和包膜的混.
- 专有LLM显示出比开源LLM更高的可复制性 (78.9%一致的单元),其中显示出单元错误和跳过的字符.
结论:
- 基于LLM的形态描述挖掘是可行的,但需要非常精确的指令.
- 与脚本相比,LLM 本质上是概率性的,对完全可复制性和集成到自动化工作流程提出了挑战.
- 未来的研究应该研究提取增强生成或微调,以提高在形态数据提取中的LLM性能.
相关概念视频
02:16Light Acquisition
8.6K
In order to produce glucose, plants need to capture sufficient light energy. Many modern plants have evolved leaves specialized for light acquisition. Leaves can be only millimeters in width or tens of meters wide, depending on the environment. Due to competition for sunlight, evolution has driven the evolution of increasingly larger leaves and taller plants, to avoid shading by their neighbors with contaminant elaboration of root architecture and mechanisms to transport water and nutrients.
8.6K
02:40Pollination and Flower Structure
68.1K
Flowers are the reproductive, seed-producing structures of angiosperms. Typically, flowers consist of sepals, petals, stamens, and carpels. Sepals and petals are the vegetative flower organs. Stamens and carpels are the reproductive organs.
68.1K
02:27Basic Plant Anatomy: Roots, Stems, and Leaves
60.5K
The primary organs of vascular plants are roots, stems, and leaves, but these structures can be highly variable, adapted for the specific needs and environment of different plant species.
60.5K
03:36Genome Annotation and Assembly
19.3K
The genome refers to all of the genetic material in an organism. It can range from a few million base pairs in microbial cells to several billion base pairs in many eukaryotic organisms. Genome assembly refers to the process of taking the DNA sequencing data and putting it all back together in a correct order to create a close representation of the original genome. This is followed by the identification of functional elements on the newly assembled genome, a process called genome annotation.
19.3K
02:54Evolutionary Relationships through Genome Comparisons
6.1K
Genome comparison is one of the excellent ways to interpret the evolutionary relationships between organisms. The basic principle of genome comparison is that if two species share a common feature, it is likely encoded by the DNA sequence conserved between both species. The advent of genome sequencing technologies in the late 20th century enabled scientists to understand the concept of conservation of domains between species and helped them to deduce evolutionary relationships across diverse...
6.1K
01:31Taxonomy
77.2K
Taxonomy is the science of defining and naming groups of biological organisms based on shared characteristics. It uses a hierarchy of increasingly inclusive categories with Latin names. The smallest units of taxonomy, species and genus, are used to assign a formal, taxonomic name to each species in a system. This classification system, referred to as binomial nomenclature, was formalized by Carolus Linnaeus in the 18th century.
Hierarchy of Taxonomy
The hierarchy that Carolus Linnaeus first...
Hierarchy of Taxonomy
The hierarchy that Carolus Linnaeus first...
77.2K