Relation Extraction From Unstructured Species Descriptions Using TaxonNERD and LLaMA 2 7B
Bibliographic record
Abstract
Ontologies are essential tools for organizing information on taxonomy, ecology, and inter-species relationships, helping to standardize ecological data and facilitate integration of large datasets. Combining ontologies with advanced Natural Language Processing (NLP) techniques, such as Named Entity Recognition (NER) and Relation Extraction (RE), has greatly improved the discovery of insights from unstructured scientific texts, particularly in biodiversity (Gabud et al. 2023, Abdelmageed et al. 2022, Hearst 1992). This study combines ontologies and NLP to analyze complex trophic interactions among animal species (Gabud et al. 2023), using a dataset (National Biodiversity Institute of Costa Rica (INBio) 2015) containing species descriptions in English and Spanish. We applied TaxoNERD to identify taxonomic entities (Le Guillarme and Thuiller 2021) and we fine-tuned the Large Language Model Meta AI (LLaMA 2 7B) to extract feeding interactions and predator-prey relationships (CheeKean 2023), due to its effectiveness in handling complex language patterns and its adaptability to diverse scientific domains. Our results (Fig. 1) showed a recall of 0.73 and a precision of 0.68, indicating that the model effectively identifies feeding relationships in most cases. However, the lower precision suggests that the model may still capture some unrelated interactions, highlighting an area for improvement to reduce false positives and increase accuracy (Touvron et al. 2023). Previous studies also emphasize the need for further refinement of relation extraction models to enhance accuracy (Mora-Cross et al. 2023). The structured dataset offers valuable insights into species’ diets and roles, contributing to biodiversity research and conservation efforts (Mora-Cross et al. 2023, Touvron et al. 2023). Moreover, this research highlights the potential of integrating AI-driven tools with ontological frameworks to manage and analyze biodiversity data at scale (Abdelmageed et al. 2022). By transforming unstructured text into structured data, we make ecological information more accessible, supporting better decision-making in conservation strategies (Abdelmageed et al. 2022, Hearst 1992). This approach scales well with the growing volume of biodiversity data, offering a more efficient and accurate method for analyzing species interactions, which are crucial for ecosystem management and endangered species protection (Gabud et al. 2023).
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.008 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.004 |
| Bibliometrics | 0.010 | 0.005 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.004 | 0.005 |
| Open science | 0.002 | 0.004 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.012 | 0.008 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".