RecSOI: recommending research directions using statements of ignorance

Can statements of what remains unknown suggest research directions?

RecSOI recommends statements of ignorance from scientific articles: passages that identify gaps, uncertainties, or unanswered questions. It matches these passages to a researcher's publication history and presents their surrounding context, helping readers investigate questions connected to their expertise.

What the study found

• The system turns explicit knowledge gaps in the literature into personalized recommendations.
• A concept-overlap evaluation found that recommended statements often connected to concepts in the researcher's abstracts.

How the study works

The method combines embeddings of statements and their context with a weighted profile built from a researcher's prior abstracts. A heuristic evaluation checks whether recommended statements share concepts with the researcher's work.

Scope and limitations

• Concept overlap does not establish that a suggested direction is novel, feasible, or scientifically productive.
• An unanswered question in an older article may have been resolved by later research.

Using this work

The linked repository provides the implementation and supporting resources. Read recommended statements in their original context and check newer literature before treating them as open questions. Consult the repository's resource and license information for your intended use.

Implementation and documentation

Cite this work

Adrien Bibal, Nourah M. Salem, Rémi Cardon, Elizabeth K. White, Daniel E. Acuna, Robin Burke, Lawrence E. Hunter (2024). RecSOI: recommending research directions using statements of ignorance. Journal of Biomedical Semantics. https://doi.org/10.1186/s13326-024-00304-3

@article{Bibal2024,
  title = {RecSOI: recommending research directions using statements of ignorance},
  author = {Bibal, Adrien and Salem, Nourah M. and Cardon, R{\'e}mi and White, Elizabeth K. and Acuna, Daniel E. and Burke, Robin and Hunter, Lawrence E.},
  year = {2024},
  publication_date = {2024-04-22},
  journal = {Journal of Biomedical Semantics},
  volume = {15},
  number = {1},
  pages = {2},
  doi = {10.1186/s13326-024-00304-3},
  url = {https://doi.org/10.1186/s13326-024-00304-3}
}