Looking for help to use AI (LLM) for a Systematic Literature Review in Soil Biodiversity/agroecology Research

Research
1

Hi HF community,

My name is Marie-Liesse Vermeire, and I am a researcher at CIRAD (the French Agricultural Research Centre for International Development). My current work focuses on assessing the impact of recycling organic waste in agriculture on soil organisms. As part of this research, I am leading a project to create an evidence map (the goal is to systematically review the literature, identify knowledge gaps, and highlight areas in need of further research or meta-analyses).

Our team includes 17 researchers from institutions across France (CIRAD and INRAe), Belgium (Gembloux-ULiĆØge), and Canada (University of British Columbia). We have gathered a research equation with approximately 27,000 articles, of which around 12,000 are potentially relevant. Given the large volume of data, we are considering using AI tools for PICO extraction and possibly metadata extraction as well. However, weā€™ve encountered challenges, as research in natural language processing (NLP) applied to agronomy and ecology is less developed compared to fields like medicine (and none of us are programmers).

We are looking for a collaborator, specialised in AI (GPT or Llama for example), that would be interested in collaborating on our project. Thank you for considering my request. I look forward to your response and am happy to provide further details if needed.

Best regards,

ML

2 Likes
2

Hello Marie i would love to collaborate on this research. i have experince in creating multimodal ai and rag application (basically extracting information form pdf and other document sources). and i also have experience in using open source language models like llama, qwen, gemma and other models as well. and i am very much intrested in collaborating

2 Likes
4

Thank you very much for your answer! Iā€™ll send you a DM

1 Like
5

Hi, I would like to collaborate on this project, too. I have been working on extracting insights from text documents using LLM and implementing RAG to improve generation quality. Is there still available to join?

2 Likes
6

Hi Marie-Liesse,

Thank you for sharing your project details. Iā€™m very interested in the opportunity to collaborate on applying AI and NLP tools to your evidence mapping project in agronomy and ecology. Although many NLP applications have been developed in the medical domain, Iā€™ve been working on adapting models like GPT and Llama for domains with less standardized terminologiesā€”and Iā€™m excited about the challenges that arise from working with ecological and agronomic texts.

A few points about how I could contribute:

  1. PICO and Metadata Extraction:
    I have experience in fine-tuning language models for information extraction tasks. We could explore creating custom pipelines to extract Population, Intervention, Comparison, and Outcome elements (or domainā€specific equivalents) from the literature. I can also help in designing systems for extracting other key metadata that would support the evidence mapping.

  2. Domain Adaptation:
    I understand that the language used in agronomy and ecology might differ from that in medical texts. Iā€™ve worked on domain-adaptive pre-training and can help create a model that is better attuned to your literature corpus.

  3. Scalability:
    With 27,000 articles (and 12,000 potentially relevant), we will need to develop a workflow that scales efficiently. I can assist with designing and implementing a pipeline (potentially using GPT or Llama-based models) that automates the extraction process, possibly incorporating active learning to refine outputs iteratively.

  4. Collaboration & Communication:
    I appreciate that not everyone in your team is a programmer. Iā€™m committed to building user-friendly tools (or providing clear documentation and training) so that the methods we develop can be used and further refined by your researchers.

I would be happy to discuss the project in more detailā€”understanding your specific requirements, the nature of your data, and your timelines. Please feel free to contact me via [your preferred contact method] or reply here to set up a meeting.

Looking forward to the possibility of collaborating on this important research.

Best regards,

Alan

2 Likes