Why this genus? Mapping genus–domain relationships in microalgae research using bibliometrics and LLMs

This study presents a data-driven mapping of the relationships between major microalgal genera (48 in total) and their associated research domains. To do so, it combines bibliometrics (corpus of 50 248 articles featuring the keyword "microalga*") and Large Language Models (LLMs). An initial semantic network↦concepts identification↦domains affectation workflow yielded perfectible results. Nevertheless allowing to identify 11 research domains structuring microalgae literature (e.g., cultivation, taxonomy, biofuel, aquafeed, …). In a second step, transitioning to a direct estimation of domain representativeness significantly improved result coherence. This refinement reduced misclassification and more effectively captured both dominant and secondary research domains for a given genus. Domain-wise normalization further clarified the relative contributions of each genus, mitigating biases arising from uneven literature coverage. Overall, the results demonstrate that, despite the inherent simplifications of bibliometric and semantic approaches, the proposed methodology reliably captures large-scale research domains and meaningful specialization patterns within the microalgae community (e.g., Haematococcus for pigment production, Chlamydomonas for photosynthesis, Microcystis for ecophysiology - algal bloom -, …). From a technical standpoint, Llama 3.3 70B Instruct was the best-performing LLM for genus-specific concept identification, while GPT OSS 120B excels at semantic network-to-research domain mapping. Remaining limitations, such as residual noise in taxonomy-related concepts and sensitivity to data representation, highlight avenues for further improvement. Nonetheless, the framework provides a synthetic and interpretable overview of how microalgal diversity is functionally distributed across scientific and applied domains. From a practical standpoint, this mapping offers a clearer perspective on the implicit rationale underlying genus selection.