Set up a pipeline to extract SGD gene names only for a list of test papers

[valearna]

Run this pipeline on caltech-curation-dev to evaluate how well the extraction works. Select a list of 20 test papers that are in corpus for WB and SGD and have C. elegans and S. Cerevisiae species tags. Store the results in a csv file with paper id and list of sgd gene ids and names

[valearna]

test_sgd_genes_extraction_results.csv

[valearna]

@draciti @vanaukenk Here's the results of the SGD genes extraction on 20 papers that are in corpus for both WB and SGD

[draciti]

thanks @valearna !
@vanaukenk and I will check those and we will get back to you.
@draciti will take the first half, @vanaukenk the second half.
Will evaluate if the genes extracted for SGD are correct, and if there was any missed also for the papers that did not have any gene extracted.

[draciti]

Moved the doc on drive: https://docs.google.com/spreadsheets/d/1-f5nHKNLPwRc0ymi-xhPdy3eQPa_LGYtTjQqodf43dE/edit?gid=572566362#gid=572566362

[draciti]

@valearna -- Are we also applying TF-IDF with a threshold of 10 for the list of extracted genes for yeast?

[valearna]

Yes, we are using TFIDF 10 also for SGD

[draciti]

@vanaukenk I am done with the first half -FYI

[vanaukenk]

I've finished checking my S. cerevisiae papers.

Generally, I thought the ACKnowledge pipeline did quite well, but one major thing I saw was that we will need to account for protein names that are different from gene names and decide how we want to find those. We missed some SGD gene associations because the mention was actually a protein, for example Puf4p instead of PUF4.