Harry Caufield
Harry Caufield
BioRED evaluation. This evaluation measures performance of OntoGPT on relation extraction over the BioRED data set (see Luo et al. 2022, https://doi.org/10.1093/bib/bbac282).
Primarily of relevance to the GPT-16k models (see also #133 ), but also to any model with a context limit >4k. Limits are currently hardcoded in may places, but shouldn't...
Extraction (SPIRES) output should include model name (including source), time, and date. It may be useful to include whether the query was run vs. live API or vs. cached result,...
Running the following command: ``` $ ontogpt -vvv pubmed-annotate -t gocam --get-pmc --limit 1 "27849154" ``` results in an error: ``` ... INFO:ontogpt.engines.knowledge_engine:GROUNDING HSN using Gene INFO:ontogpt.engines.knowledge_engine: Annotators: ['gilda:', 'sqlite:obo:pr']...
Pydantic versions of the extraction templates are generated through linkml's `gen-pydantic`, and that's currently called through the Makefile doing this: ``` poetry run gen-pydantic --pydantic_version 2 $< > $@ ```...
This would apply for all SPIRES extractions, but particularly for PubMed extractions. It would be useful to know summary statistics, e.g., how many entities, how many relations, how many grounded...
@enockniyonkuru suggests including a threshold on PubMed extraction outputs, such that we can select for documents yielding the same entities or relations (e.g., specify which results appear in 3 or...
It's a bit awkward to pass pre-compiled corpora of documents (as lists of PMIDs) to the PubMed-based extractor. It could be easier just to provide an input file with these...
Several releases (like https://github.com/monarch-initiative/ontogpt/releases/tag/v0.3.0) have examples in the notes. These examples aren't available elsewhere, so including them in the documentation would be very helpful.
Calls to the `recipe-extract` command fail with the following: ``` Traceback (most recent call last): File "/home/harry/ontogpt/.venv/bin/ontogpt", line 6, in sys.exit(main()) File "/home/harry/ontogpt/.venv/lib/python3.9/site-packages/click/core.py", line 1157, in __call__ return self.main(*args, **kwargs)...